mirror of
https://github.com/coder/coder.git
synced 2026-10-31 19:36:41 +08:00
9ce42293b0a760c80f728ad184a00e83e5d53ffe
4412
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 david-fraley
|
a2afb458c5 |
feat: add premium paywall conversion telemetry (#28425)
**TLDR:** When someone using free Coder taps a "Start trial for free" button, we now write down which page they tapped it on, and whether they went on to actually start the trial. That tells us which locked features make people want Premium. ## Brief summary - Records two things: the button being clicked, and a trial actually being started. - Each click gets a random ID that travels with the person to the trial form, so we can tell who finished from who walked away. - Records a short label for the page (like `appearance` or `audit_log`), never the web address, so no customer names are collected. #28226 added the trial form and has since merged, so this now targets `main`. ## Description Today the "Start trial for free" button just moves you to the Premium page and nothing is recorded, so there is no way to know which feature convinced someone. This adds two events: | Event | Happens when | |---|---| | `cta_click` | Someone clicks "Start trial for free" | | `trial_signup` | A trial licence is actually issued | The click generates a random ID. It rides along to the trial form and gets saved on the signup, so a click and a signup can be matched up later. Clicks with no matching signup are the people who gave up. Trials started without a paywall (for example straight from the sidebar) are labelled `direct`, so "no label" never gets confused with "broken". A few small choices worth knowing: - The ID is kept in browser session storage, not in the web address, so the URL stays clean. It expires after 30 minutes. - The trial signup event is recorded by the server, not the browser, so nobody can fake a signup. - The browser cannot say which kind of event it is sending. It can only report a click; the server labels it. - Only people allowed to start a trial are counted. The server rejects the rest, which matters because a few paywalls show their button to everyone. - Nothing is recorded at all when telemetry is turned off. There is deliberately no "paywall was shown" event. Every admin page visit would report one, which is a lot of noise for little insight. Every paywall in the [Premium Copy Updates](https://www.notion.so/coderhq/Premium-Copy-Updates-3c1d579be59280a59557cabd54fd8a1f) table has a label. The rows that do not are badges and alerts with no button to click. ## Proof Ran a local Coder with no licence and clicked the paywall button on the Appearance page. What the browser sent, and the reply: ```json POST /api/v2/deployment/premium-funnel-events -> 204 No Content {"id":"92d96afa-f975-4789-b8a7-b1fd19d4df85","source":"appearance","variant":"premium"} ``` What the browser saved for the trial form, same ID: ```json {"id":"92d96afa-f975-4789-b8a7-b1fd19d4df85","source":"appearance","createdAt":1787268520857} ``` The click then landed on the Premium page with the trial form. Screenshots of the network panel, the saved value, and the Premium page are attached below. The form itself was not submitted, since that would contact the real licence server. ## Testing - Backend tests: a click is recorded with the server's own label, an unknown page label is refused, and a normal member is refused. - Storybook tests: button clicked and ID saved on all three paywall styles, plus the case where the viewer only sees guidance text and has nothing to click. - Unit tests for the saved ID: expiry, bad data, unknown page label, and the `direct` fallback. - `make gen` is clean, and lint, types, and formatting pass. Two stories fail on `main` today and fail the same way here: `ExternalAuthSettingsPageView > Page` and `PremiumPageView > No License`. Neither file is touched by this branch; both were verified failing on `main` with this branch checked out elsewhere. ## Follow-up coder/coder-telemetry-server#45 saves these events for reporting. It merges after this one. <details> <summary>Plan and decisions</summary> **Goal:** find out which Premium paywall drives trial signups, and tell abandons apart from completions. **Options considered and dropped:** 1. Put `?from=appearance` in the URL. Rejected: messy for users. 2. Use router state. Rejected: lost if the page is refreshed. 3. Let the browser report the signup. Rejected: it could be faked or lost during the redirect. 4. Work out the page from the URL. Rejected: URLs contain organization and template names. 5. Report every paywall impression. Rejected: one event per admin page visit is noise, and the click is the signal. 6. Separate event types and tables. Rejected: makes the matching query harder for no gain. **Event shape:** ```go type PremiumFunnelEvent struct { ID uuid.UUID // event ID; for a click this is also the matching ID EventType string // cta_click | trial_signup, set by the server Source string // appearance, audit_log, ..., direct Variant string // which paywall design was shown AttributionID uuid.UUID // the click this signup came from UserID uuid.UUID CreatedAt time.Time } ``` **Where the code lives:** nothing under `components/` imports from `modules/`, so the paywall components stay presentational and only gain an `onCTAClick` prop. New wrappers in `modules/paywall/` own the reporting, one per paywall style, each taking a fixed `source`. **Self-review:** the `frontend-review` skill caught that reading permissions inside the wrapper forced an auth context into presentational page views, which broke their stories. Every call site already passes `canViewPremium={permissions.viewAllLicenses}` and the button only renders when that is true, so the wrapper no longer reads permissions and the server stays the gate. It also flagged one unattributed paywall on the Security page, now wired up as `browser_only`. </details> > Generated with [Coder Agents](https://coder.com/agents) on behalf of @david-fraley |
||
|
Samuel VolinandMatt Vollmer
|
f44a3b59d7 |
feat: Premium Page CTA updates and form handling (#28226)
* Premium page has form for signing up for a trial * Input validation and normalization for trial form fields * api endpoint definition to make this processes seamless * license page updates to access `?success=true` queryparam and confetti <img width="1512" height="833" alt="Screenshot 2026-08-17 at 10 08 41 PM" src="https://github.com/user-attachments/assets/1766e724-c63a-4da9-b2bd-a98e6b66cc9b" /> <img width="1510" height="874" alt="Aug-17-2026 22-12-38" src="https://github.com/user-attachments/assets/3a2a0f35-1176-4290-aa0d-b3f7d66c15d6" /> ```mermaid sequenceDiagram autonumber actor Owner participant UI as UI<br/>TrialRequestForm + PremiumPage participant API as API<br/>api.ts + coderd middleware participant Coderd as Coderd<br/>postTrialLicense + trialer participant Licensor as Licensor<br/>v2-licensor + Postgres + pubsub + cache Note over UI: mutation.status = "idle" Owner->>UI: fills 10 fields, checks acknowledgement UI->>UI: Yup validate, strip "acknowledged", submit UI->>API: POST /api/v2/licenses/trial<br/>cookie + X-CSRF-TOKEN Note over UI: mutation.status = "pending" API->>API: verify CSRF, apiRateLimiter,<br/>apiKeyMiddleware sets actor API->>Coderd: dispatch Coderd->>Coderd: audit.InitRequest, authorize license:create,<br/>httpapi.Read validates, HasLicense is false Coderd->>Licensor: POST trial request<br/>deployment_id unspoofable, 10s deadline Licensor-->>Coderd: 200, raw signed JWT Coderd->>Coderd: ParseClaimsIgnoreNbf, uuid.Parse Coderd->>Licensor: InsertLicense, updateEntitlements,<br/>publish PubsubEventLicense, commit audit log Licensor-->>Coderd: database.License Coderd-->>API: 201 codersdk.License API-->>UI: 201 codersdk.License Note over UI: mutation.status = "success" UI->>Licensor: invalidate entitlements + licenses keys UI->>Owner: navigate /deployment/premium?success=true Note over Owner: confetti fires on the success param ``` --------- Co-authored-by: Matt Vollmer <matthewjvollmer@outlook.com> |
||
|
Ehab Younes
|
903d7b756d |
feat: add workspace agent session count protocol (#28125)
Phase 2 of the Scalable Approach for Adding New IDE Session Types RFC, tracked in #27410. Adds a session_counts map to Stats on the Agent API, so an agent can report a count per app name instead of the four fixed fields, which are now marked deprecated. Bumps the API to v2.11. - coderd normalizes names on ingest, caps distinct entries, and sums the overflow under unknown, so an arbitrary client name cannot blow up cardinality. - Agents predating v2.11 set the fixed fields instead, never both, and the ingest path falls back to them. Nothing sends session_counts yet. The agent starts populating it in #28338. |
||
|
Michael Suchacz
|
9168dbf4d3 |
refactor(coderd/x/chatd): unify LLM model-call construction behind a resolver (#28145)
Centralizes every LLM client build in chatd behind one resolver and makes all model calls behave the same, instead of preserving per-flow quirks. ## Why `chatd` built LLM clients in five places (standard turn, compaction override, quickgen, title override, advisor), each repeating config parsing, route resolution, client construction, and debug wrapping with slightly different behavior. An earlier iteration of this PR centralized construction but preserved every historical quirk behind purpose-specific spec constructors and policy flags. Review concluded the quirks themselves were the problem. ## What changed - `Server.resolveModelCall` (`modelcall.go`) is the single pipeline from a `modelCallSpec` to a ready client plus call metadata. Callers build spec literals inline; the only spec fields are data: model source (chat config, explicit config row, or fixed provider/model pair for computer use), requested reasoning effort, chatd-scoped routing, and build options. - Uniform behavior for every call: - Provider options are always derived. Whole-chat summaries, turn status labels, and chat-model compaction summaries previously sent none; they now follow the config like everything else. - `MaxOutputTokens` always defaults to 32k when the config leaves it unset. The compaction summary call nulls the cap after copying the template because it is non-streaming and the Anthropic SDK rejects non-streaming requests whose cap implies a completion over ten minutes. - Debug recording always wraps the client when chat debug is enabled, labeled with the resolved provider/model. The debug-rebuild paths (`manualTitleDebugSpec`, `quickgenDebugSpec`) are deleted; debug helpers now only create and finalize run records, since run association flows through the generate context, not the model. - Clients are always constructed with the configured model string and unmutated route, so AI Gateway validation sees the name exactly as configured. The resolved identity feeds metadata, logs, and debug labels only. - `resolvedModelCall.newCall` / `newObjectCall` are the only production constructors of `fantasy.Call` / `fantasy.ObjectCall`; object calls now carry the config's sampling parameters. - Net: -240 lines. ## Validation - Full `./coderd/x/chatd/...` suite green (PostgreSQL-backed). - Red-green toggles verified for the three key unifications (option derivation, token default, summary cap removal). - ARCHITECTURE.md: prose the refactor made stale is deleted; no new resolver documentation is added. > Mux acted on Mike's behalf for this PR. |
||
|
Bobby Ho
|
6348c83142 |
feat: register public clients without a secret (#28046)
Third in the stack, split out of #27873. This is where dynamic client registration first produces a public client. - A registration requesting token_endpoint_auth_method `none` issues no secret and persists the client_type derived in #28043. The response omits client_secret entirely, per RFC 7591 §3.2.1. - Discovery still advertises only the two secret-based methods. `none` is withheld until the token endpoint honors it, so a conforming client is not told to attempt an exchange that would be rejected. - The app and its secret are written in one transaction. They were two independent inserts, so a failed second one left a committed app that could never authenticate while still holding a registration access token. Pre-existing, but a public client's missing secret row makes the orphaned case unspottable by inspection. - registration_client_uri uses url.JoinPath, fixing a trailing-slash access URL producing //oauth2/clients/{id}. - Docs state both limits above, so the page does not describe a flow that returns 400 until #28047 lands. A client registered here cannot obtain a token yet; #28047 adds that. Dynamic client registration is off by default, so none of this is user-visible until then. Refs https://linear.app/codercom/issue/ENG-3029/oauth2-support-public-client |
||
|
Yevhenii Shcherbina
|
211f490910 |
refactor: share the AI Gateway price book contract (#28310)
Follow-up to the review discussion on the price book refresh automation: https://github.com/coder/coder/pull/28146#discussion_r3806439966 ## Problem The price seed row struct was duplicated three times, so a JSON tag change had to be repeated in each copy: - `scripts/aibridgepricesgen.priceRow` - `scripts/aibridgepricesdiff.priceRow` - `coderd/aibridge/prices.seedRow` The `(provider, model)` key and its parsing were duplicated too. ## Change Add `coderd/aibridge/prices/pricebook` holding the shared contract: `Row`, `Key`, `Row.Key()`, and `Parse`. The package depends only on the standard library, so the scripts can import it without pulling in the database layer that `coderd/aibridge/prices` brings. Consumer-specific behaviour stays where it is: upstream conversion and validation in the generator, comparison and Markdown rendering in the diff tool, DB seeding and runtime lookup in the seeder. The batch SQL upsert still extracts JSON fields by name, so it remains a separate contract boundary. The note about keeping it in sync now lives on the shared `Row`. No generated artifacts change; the serialized schema is identical. --- <sub>Opened by Coder Agents on behalf of @evgeniy-scherbina.</sub> |
||
|
Bobby HoandClaude Opus 5
|
dea3a2944a |
feat(coderd/rbac): compare scopes by permission coverage (#28167)
Split out of #28045 ([PLAT-479](https://linear.app/codercom/issue/PLAT-479)) so the coverage rules can be reviewed on their own. First of three. No behavior change: both functions are added without production callers, and the OAuth2 authorize endpoint consumes them in the next PR. **ScopesCover(allowed, requested)** reports whether every permission the request grants is also granted by the allowlist. It expands both sides and compares permissions rather than names, so coder:workspaces.access covers workspace:read without ever naming it. The fail-closed rules are the part worth reviewing: - Coverage models site-level grants only. A scope carrying org or user permissions, a negative permission, or a narrowed allow list is refused on either side rather than compared on the part that is modeled. Answering "covered" from the fraction that was read would report authority that was never examined. - An unknown name on either side is an error, not a false, since a caller cannot tell "not covered" apart from "could not decide". - A wildcard request is covered only by a wildcard grant. Enumerating every workspace action that exists today does not cover workspace:*, because the wildcard also authorizes the actions added tomorrow. **CanonicalScopeName** maps the aliases IsExternalScope accepts, all and application_connect, onto the api_key_scope enum spellings. IsExternalScope answers whether a name may be requested, not how it is spelled once persisted, so anything storing a validated name has to canonicalize in between. Both sides of ScopesCover must already be canonical, which its parameter names restate at every call site. Covered by 17 coverage cases, guard cases that drive each unmodeled shape through both sides of the comparison, and an exhaustive sweep over the external catalog: coder:all covers all of it, and every catalog name covers itself. Stack: this PR, then negotiate and persist the scope, then the consent page and invalid_scope redirect. --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com> |
||
|
Michael Suchacz
|
8eb1e43f0b |
feat: apply reasoning effort and surface thinking blocks for Google chat models (#28273)
## Problem
Selecting a reasoning effort on a Google (Gemini) chat model either got
dropped or broke the request outright, and even when the request
succeeded, Gemini chats never showed thinking blocks:
- The native `fantasygoogle` path had no case in `applyReasoningEffort`
at all, so effort was silently ignored. With only `include_thoughts`
configured, Gemini 3+ models default to no thinking, so chats produced
zero reasoning tokens.
- The path chat generation actually uses today routes Google providers
through aibridge as openai-compat clients and forwarded the raw effort
string as `reasoning_effort`. Google's OpenAI-compatible endpoint
validates that value against the model's supported thinking levels
instead of clamping, so out-of-range efforts fail the whole request with
HTTP 400. Verified against the live Gemini API: `gemini-3.1-pro-preview`
accepts low/medium/high and rejects none/minimal/xhigh/max. Because the
config's default effort is sent whenever the user selects nothing, a
config whose default falls outside the model's supported set failed
every no-selection message.
- Google's OpenAI-compatible endpoint never emits thought text for
`reasoning_effort` requests. Thoughts require
`extra_body.google.thinking_config` with `include_thoughts`, which is
mutually exclusive with `reasoning_effort`, and even then the thought
deltas come back inline in `content` (marked
`extra_content.google.thought`, wrapped in `<thought>` markers) rather
than in the `reasoning_content` field OpenAI-compatible clients parse.
So no reasoning parts were ever produced, persisted, or rendered for
Gemini chats (verified live in UAT).
## Fix
Reasoning effort:
- Map the global reasoning effort scale onto Google's `thinking_level`
in `applyReasoningEffort`, gated to Gemini 3+ models and clamped to each
model's supported subset (3.0 Pro launched with LOW/HIGH only, 3.1 Pro
added MEDIUM, Flash accepts all four through 3.6 while 3.7 and the
latest flash aliases drop MINIMAL, image models accept HIGH plus MINIMAL
for flash). Pre-Gemini-3 models reject `thinking_level`, so effort
degrades to a no-op for them and a config-pinned level is dropped at
call time instead of failing every generation.
- Add `thinking_level` to `ChatModelGoogleThinkingConfig` so admins can
pin a level for configs without user-selectable effort. A resolved
per-turn effort overrides the pinned level; an explicit
`thinking_budget` wins over both since the Google API rejects requests
that set `thinking_budget` and `thinking_level` together, and config
validation now rejects that combination up front. The admin panel picks
up the new field automatically through the generated model-options
schema.
Thinking blocks on the openai-compat path (the path chat generation
uses), translated in both directions at the existing Gemini transport
seam:
- chatd's OpenAI-compat request patch swaps `reasoning_effort` for an
equivalent `extra_body.google.thinking_config` with `include_thoughts`
enabled: `thinking_level` on Gemini 3+ (clamped to the model's supported
subset), `thinking_budget` on earlier models (effort `none` maps to
budget 0, which disables thinking). Non-Gemini models and non-Google
routes are untouched.
- The same transport rewrites Google's responses so thought output
surfaces as `reasoning_content`, which fantasy already converts into
reasoning parts: streaming bodies are rewritten per SSE line (preserving
streaming latency), JSON bodies in place, with the
`<thought>`/`</thought>` markers stripped at the transitions.
- aibridge's chat-completions interception preserves the `extra_body`
passthrough that openai-go's typed params silently drop, forwarding it
to Google upstreams only, so the chatd -> aibridge -> Google route
carries the thinking config end to end.
## Validation
- `go test ./coderd/x/chatd/... ./aibridge/... ./internal/googleopenai
./codersdk` pass; golangci-lint clean on touched packages.
- New unit tests cover the effort-to-level mapping for every effort
value, per-model clamping on both paths, the effort-to-thinking_config
request swap (level, budget, pinned-config precedence, non-Gemini
passthrough), SSE and JSON thought-response rewriting (fixtures mirror
live captures, including tool-call and no-thought streams), an
end-to-end fake-SSE test through the fantasy pipeline proving reasoning
deltas emerge with markers stripped, and aibridge `extra_body`
preservation (including proof the typed round trip drops it).
- Verified live against `generativelanguage.googleapis.com` on
`gemini-3-flash-preview` and `gemini-2.5-flash` through the real
chatprovider stack: reasoning deltas stream through with markers
stripped and answer text intact.
- Live UAT on a dogfood deployment (remote Coder Agents chat, 4 rounds):
effort clamping verified end to end (rounds 1-2), the missing-thoughts
limitation isolated (round 3), and the fix validated (round 4):
expandable, readable thinking blocks render for
`gemini-3.1-pro-preview`, `gemini-3-flash-preview`, and
`gemini-2.5-flash` at effort=high; no `<thought>` marker leaks;
effort=none produces no block and no 400; multi-turn history replay
works with a thinking block on each turn.
- `make gen` regenerated `typesGenerated.ts` and
`chatModelOptionsGenerated.json`; `pnpm -C site lint:types` passes.
> 🤖 Mux acted on Mike's behalf to create this pull request.
<!-- mux-attribution: model=claude-opus-4-6 thinking=high -->
|
||
|
Danielle Maywood
|
5480299cef |
fix: serialize empty MCP tool properties as {} not null for OpenAI (#28380)
## Summary
Fixed a bug where MCP tools with an empty `properties` schema had their
tool definitions sent to OpenAI with `"properties": null`, causing
OpenAI to reject the request.
## Problem
When an MCP server reports a tool with an empty input schema, e.g.:
```json
{
"type": "object",
"properties": {},
"required": []
}
```
Coder was forwarding the tool definition to OpenAI as:
```json
{
"type": "object",
"properties": null
}
```
OpenAI's function-calling API rejects `properties: null`, making these
tools unusable with OpenAI models.
## Root cause
The empty `properties` object collapsed to Go `nil` at two layers:
1. `agent/x/agentmcp/manager.go` `toolInputSchemaMap` stripped the
present-but-empty `"properties": {}` (guarded by `len(properties) > 0`),
so the wire copy lost it entirely.
2. `coderd/x/chatd/chatloop/chatloop.go` `buildToolDefinitions` embedded
`info.Parameters` directly into the input schema. When nil,
`encoding/json` marshals it as `null`.
## Fix
- `buildToolDefinitions` now substitutes an empty object when
`Parameters` is nil, mirroring the existing `required` guard (the sink,
protects all tool sources).
- `toolInputSchemaMap` now preserves an empty `properties` object
instead of dropping it (the upstream producer).
## Tests
Added regression tests (failing before, passing after) covering both
layers and asserting the schema serializes to
`{"type":"object","properties":{}}`:
- `TestBuildToolDefinitionsNilPropertiesBecomesEmptyObject`
- `TestToolInputSchemaMapPreservesEmptyProperties`
-
`TestWorkspaceMCPToolInfosFromResources/EmptyPropertiesYieldsEmptySchema`
---
*This PR was generated by Coder Agents.*
|
||
|
Susana Ferreira
|
363d7fc4d9 |
feat: add CTAs to AI budget admin notifications (#28138)
## Problem
AI budget admin notifications tell an admin that a user crossed a budget
threshold, but give them nothing to act on: no link to the user's
budget, and no pointer to the docs.
## Changes
- Add a **Manage user AI budget** action to both admin templates,
linking to the effective group's members page filtered to the affected
user, where the per-user override is set.
- Link to the AI budget documentation from both notification bodies.
- Say when the block lifts in the limit-reached notification ("blocked
until `{{period_end}}`").
- Add an `organization_name` label, needed to build the members page
URL.
- Grant the AI Gateway daemon subject read access to organizations,
which the label lookup requires.
- Add an authorization test for the notification path, which also covers
the existing group read grant.
<details>
<summary>Screenshots</summary>
**Admin Notifications in Coder UI**
<img width="456" height="695" alt="admin_notifications"
src="https://github.com/user-attachments/assets/a579f91c-55cc-497c-87f4-235f15493725"
/>
**Admin Notification in Email**
<img width="660" height="399" alt="admin_notifications_email"
src="https://github.com/user-attachments/assets/39b876fb-f512-41e5-997f-a5269cd26ddc"
/>
</details>
Related to internal Slack threads:
* https://codercom.slack.com/archives/C096PFVBZKN/p1786051630688569
* https://codercom.slack.com/archives/C04M9CNGHGF/p1786453662306149
> [!NOTE]
> Initially generated by Claude Opus 5, modified and reviewed by
@ssncferreira
|
||
|
Susana Ferreira
|
50ecd6959f |
feat: add a source filter to the AI model price list (#28329)
## Description With prices from the price book and prices set through the API both in the table, `list` showed only the price in effect and gave no way to tell where it came from. There was also no way to see the price book's entry for a model that had been overridden, since the listing resolves to the custom price. This adds a `source` column to the output and a `--source` filter for narrowing to one source, or to every row with `all`. Note: `--source` selects rows rather than models. A model carrying both a price book price and a custom one appears under either value, so the two filtered listings do not sum to the unfiltered one. `--source all` reports both rows at once. ## Changes - Add `Source` to `codersdk.AIModelPrice` and `AIModelPricesFilter`, and map it in `db2sdk`, which previously dropped it. - Filter `GetAIModelPrices` by source before resolution, so the price book's row for an overridden model stays reachable. - Accept `all` as a source, which reports every price a model holds instead of resolving to one. - Accept a `source` query parameter on the list endpoint, rejected with a 400 when outside the enum. - Add `--source` to the list command and a `source` column to its output. - Document the column and the filter, and lead the price section with `coder exp ai-model-prices list`. - Distinguish a deployment's default prices from the price book that ships with each release. Closes https://linear.app/codercom/issue/AIGOV-593 > [!NOTE] > Initially generated by Claude Opus 5, modified and reviewed by @ssncferreira |
||
|
Danielle Maywood
|
4a7a0a3348 | feat: label process_output rows with the process command (#28300) | ||
|
Susana Ferreira
|
19d1ee9fc8 |
feat: support custom prices for models in the price book (#28328)
## Description The experimental CLI could only price models the price book does not cover, so an operator with a negotiated rate for a model Coder already prices had no way to record it. The restriction existed because a single row served both the price book and prices set through the API, and the startup seeder would overwrite an override on the next restart. Each source now keeps its own row (https://github.com/coder/coder/pull/28204), so that no longer applies. This removes the restriction: a price set through the CLI takes effect over the price book for any model, and the book stays underneath it. ## Changes - Remove the validation that rejected models the price book covers. - Delete `prices.IsDefaultPriced` and the index behind it, which existed only for that check. - Generalize the `coder exp ai-model-prices update` help. - Document that a custom price overrides the price book and does not follow its updates. - Rework the spend approximation callout around what a custom price can and cannot correct. Closes https://linear.app/codercom/issue/AIGOV-594 > [!NOTE] > Initially generated by Claude Opus 5, modified and reviewed by @ssncferreira |
||
|
Susana Ferreira
|
9390e51422 |
feat: distinguish default and custom AI model prices (#28204)
## Description Prices in `ai_model_prices` come from two places: the price book embedded in each Coder release, and prices set through the experimental CLI (https://github.com/coder/coder/pull/27926). Until now, they were indistinguishable, so the startup seeder re-applied the book over every row, and a price someone set would silently revert on the next restart. This adds a `source` column with values `default` and `custom`, and keeps a row for each. Readers resolve a model to its custom price when one exists and fall back to the book, including the cost path that prices an interception. Keeping both rows means the price book stays current underneath an override, so a future delete can restore it. Note: the endpoint still only accepts models the price book does not cover. Overriding a price the book already carries is handled by the follow-up issue https://linear.app/codercom/issue/AIGOV-594. ## Changes - Add a migration for the `ai_model_price_source` enum and a `source` column, backfilled to `default` and then `NOT NULL`. `source` joins the primary key, so a model can hold one row per source. - Take a `source` parameter on `UpsertAIModelPrices`. Each source keeps its own row, so the price book and a custom price never overwrite each other. - Resolve to the custom price in `GetAIModelPriceByProviderModel` and `GetAIModelPrices`, so both the interception cost snapshot and the list endpoint use it. - Seed the price book as `default`, and write prices set through the endpoint as `custom`. Closes https://linear.app/codercom/issue/AIGOV-592 > [!NOTE] > Initially generated by Claude Opus 5, modified and reviewed by @ssncferreira |
||
|
cdrciandgithub-actions[bot]
|
5e0710de92 |
chore: refresh AI model price book (#28365)
## Price book changes 61 models added, 14 models removed, 51 models changed. <details> <summary>Added</summary> - azure/claude-mythos-5 - azure/claude-opus-4-7 - azure/kimi-k2.7-code - bedrock/global.openai.gpt-5.6-luna - bedrock/global.openai.gpt-5.6-sol - bedrock/global.openai.gpt-5.6-terra - bedrock/xai.grok-4.6 - copilot/gemini-3.7-flash - copilot/grok-4.6 - copilot/kimi-k3 - copilot/mai-code-1.1-flash - google/gemini-3.7-flash - openrouter/bytedance-seed/seed-2-1-turbo - openrouter/bytedance-seed/seed-2.0-code - openrouter/deepseek/deepseek-v4-pro-0813 - openrouter/dots-studio/dots-3-note-preview:free - openrouter/google/gemini-3.7-flash - openrouter/inclusionai/ling-3.0-flash - openrouter/liquid/lfm-2.5-2.6b:free - openrouter/meta/muse-glimmer-30b - openrouter/meta/muse-spark-1.2 - openrouter/nvidia/nemotron-3.5-lightning - openrouter/nvidia/nemotron-3.5-lightning:free - openrouter/qwen/qwen3.8-2.4t-a95b - openrouter/qwen/qwen3.8-27b - openrouter/sakana/sakana-namazu - openrouter/upstage/solar-pro4 - openrouter/x-ai/grok-4.6 - openrouter/z-ai/glm-5.2:free - openrouter/z-ai/glm-5.3 - openrouter/~z-ai/glm-latest - vercel/alibaba/qwen3.8-2.4t-a95b - vercel/alibaba/qwen3.8-27b - vercel/deepseek/deepseek-v4-pro-0813 - vercel/google/gemini-3.7-flash - vercel/inclusionai/ling-3.0-flash - vercel/meta/muse-glimmer-30b - vercel/meta/muse-spark-1.2 - vercel/meta/muse-spark-1.2-contributor - vercel/nvidia/nemotron-3.5-lightning - vercel/openai/gpt-4.1-fast - vercel/openai/gpt-4.1-mini-fast - vercel/openai/gpt-4.1-nano-fast - vercel/openai/gpt-4o-fast - vercel/openai/gpt-4o-mini-fast - vercel/openai/gpt-5-fast - vercel/openai/gpt-5-mini-fast - vercel/openai/gpt-5.1-thinking-fast - vercel/openai/gpt-5.2-fast - vercel/openai/gpt-5.3-codex-fast - vercel/openai/gpt-5.4-fast - vercel/openai/gpt-5.4-mini-fast - vercel/openai/gpt-5.5-fast - vercel/openai/gpt-5.6-luna-fast - vercel/openai/gpt-5.6-sol-fast - vercel/openai/gpt-5.6-terra-fast - vercel/openai/o3-fast - vercel/openai/o4-mini-fast - vercel/sakana/namazu - vercel/xai/grok-4.6 - vercel/zai/glm-5.3 </details> <details> <summary>Removed</summary> - anthropic/claude-opus-4-1 - anthropic/claude-opus-4-1-20250805 - google/gemini-2.0-flash - google/gemini-2.0-flash-lite - google/gemini-3-pro-preview - openrouter/ai21/jamba-large-1.7 - openrouter/inclusionai/ling-3.0-flash:free - openrouter/mancer/weaver - openrouter/openai/gpt-5.3-chat - vercel/anthropic/claude-opus-4.1 - vercel/inclusionai/ling-3.0-flash-free - vercel/openai/gpt-5.1-instant - vercel/openai/gpt-5.3-chat - vercel/zai/glm-4.6v </details> <details> <summary>Changed</summary> - azure/gpt-5.6-luna - azure/gpt-5.6-sol - azure/gpt-5.6-terra - bedrock/openai.gpt-5.6-sol - openrouter/deepseek/deepseek-chat-v3-0324 - openrouter/deepseek/deepseek-v3.1-terminus - openrouter/deepseek/deepseek-v4-flash - openrouter/deepseek/deepseek-v4-flash-0731 - openrouter/deepseek/deepseek-v4-pro - openrouter/google/gemini-3.6-flash - openrouter/google/gemma-4-31b-it - openrouter/gryphe/mythomax-l2-13b - openrouter/minimax/minimax-m2.5 - openrouter/minimax/minimax-m2.7 - openrouter/mistralai/mistral-small-3.2-24b-instruct - openrouter/moonshotai/kimi-k2.5 - openrouter/moonshotai/kimi-k2.6 - openrouter/moonshotai/kimi-k2.7-code - openrouter/openai/gpt-5.6-luna - openrouter/openai/gpt-5.6-luna-pro - openrouter/openai/gpt-5.6-sol - openrouter/openai/gpt-5.6-sol-pro - openrouter/openai/gpt-5.6-terra - openrouter/openai/gpt-5.6-terra-pro - openrouter/openai/gpt-oss-120b - openrouter/qwen/qwen-plus-2025-07-28:thinking - openrouter/qwen/qwen2.5-vl-72b-instruct - openrouter/qwen/qwen3-14b - openrouter/qwen/qwen3-235b-a22b-2507 - openrouter/qwen/qwen3-30b-a3b - openrouter/qwen/qwen3-coder-30b-a3b-instruct - openrouter/qwen/qwen3-vl-235b-a22b-thinking - openrouter/qwen/qwen3-vl-30b-a3b-instruct - openrouter/qwen/qwen3.5-35b-a3b - openrouter/qwen/qwen3.6-27b - openrouter/qwen/qwen3.6-35b-a3b - openrouter/tencent/hy3-preview - openrouter/thinkingmachines/inkling - openrouter/thinkingmachines/inkling-small - openrouter/z-ai/glm-5 - openrouter/z-ai/glm-5.2 - openrouter/~deepseek/deepseek-v4-flash-latest - openrouter/~google/gemini-flash-latest - openrouter/~moonshotai/kimi-latest - openrouter/~openai/gpt-latest - openrouter/~x-ai/grok-latest - vercel/deepseek/deepseek-v4-flash - vercel/deepseek/deepseek-v4-pro - vercel/google/gemini-3.6-flash - vercel/openai/gpt-5.6-sol - vercel/zai/glm-5.2 </details> ## Review notes Regenerated by `make gen/aibridge-prices` from the live [models.dev](https://models.dev) catalog. Both artifacts come from one snapshot, so they ship together: - `coderd/aibridge/prices/data/prices.json` - `site/src/pages/AgentsPage/components/ChatModelAdminPanel/knownModels/knownModelsGenerated.json` These are customer-visible cost numbers taken from upstream data, so this PR is never merged automatically. The summary above lists what moved; check the diff for exact figures before approving. Opened automatically by the [aigateway-prices-refresh workflow](https://github.com/coder/coder/actions/runs/32352560765). Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> |
||
|
Spike Curtis
|
f5d42a868e |
refactor(coderd): add related-data selection to workspace queries (#28302)
This is the first of three PRs implementing lite `codersdk.Workspace` responses ([GRU-82](https://linear.app/codercom/issue/GRU-82), [RFC](https://app.notion.com/p/coderhq/Lite-Workspace-API-Requests-3aad579be5928060b5e7cc65c539717f)). A fully populated workspace fans out into a large tree of DB queries (template, latest build, provisioner job + queue position, resources, metadata, agents, apps, statuses, scripts, log sources, template version). Most callers do not need most of that data, and one of these queries, `GetProvisionerJobsByIDsWithQueuePosition`, is consistently the top resource consumer in scale tests. This PR adds the internal plumbing to load only a selected subset of that tree. It introduces a `workspaceRelated` selection type and threads it through `workspaceData` and `workspaceBuildsData` down to the individual query call sites, skipping any query whose data was not selected. The selection is a pointer tree that mirrors the parent/child relationships between the objects: branch nodes are pointers (non-nil when selected) and leaf nodes are bools. Modeling it this way makes selecting a child without its parent unrepresentable, which matches how loading works (a parent must be queried to learn its children's identifiers), so there is no normalization step to forget. No behavior change: every existing caller passes a fully populated selection (`allWorkspaceRelated` / `allLatestBuildRelated`), so responses are identical. Wiring the HTTP `include_related` query parameter (Phase 2) and narrowing individual callers (Phase 3) come in follow-up PRs. Two notes on the current mapping: - `queue_position` is modeled as its own node, but only one job query exists and it always computes the ranking, so selecting `job` runs the full query today. Splitting off a cheaper queue-position-free job query is deferred (it needs a new SQL query + `make gen`). - The workspace-level `LatestAppStatus` is gated on the `latest_build.resources.agents.apps.statuses` node. It is queried directly by workspace ID, but a caller selecting it also pulls in the build chain as ancestors. Happy to decouple this if preferred. <details> <summary>Implementation plan</summary> # Lite Workspace API — Phase 1 Plan Add an internal configuration structure to the Coderd method that fetches a workspace with its related data (`workspaceData` /`workspaceBuildsData`), thread it down to the SQL query call sites, and skip ("squash") queries that are not requested. Server-internal only. No HTTP query-string parsing (Phase 2) and no `wsrelated` struct tags or field nil-ing (Phase 3). ## Scope - In: config struct mapping the RFC related-data hierarchy; threading through `workspaceData` and `workspaceBuildsData`; gating every related-data query; comprehensive unit tests with a mocked `database.Store` (gomock/dbmock). - Out: HTTP `include_related` parsing, codersdk changes, `wsrelated` tags, docs autogen, and changing `convertWorkspaces` skip-on-missing-data behavior. All existing callers pass an "include all" config, so Phase 1 is a behavior- preserving refactor. ## Design New file `coderd/wsrelateddata.go` with a pointer tree that mirrors the RFC hierarchy. Branch nodes are pointers (present when non-nil); leaf nodes are bools. Because a child cannot be expressed without its parent, the RFC "includes all ancestors" rule is a structural invariant of the type: an illegal selection is unrepresentable, so no normalization pass exists or is needed. ```go type workspaceRelated struct { Template bool LatestBuild *latestBuildRelated } type latestBuildRelated struct { Job *jobRelated Resources *resourcesRelated TemplateVersion bool } type jobRelated struct{ QueuePosition bool } type resourcesRelated struct { Metadata bool Agents *agentsRelated } type agentsRelated struct { Apps *appsRelated Scripts bool LogSources bool } type appsRelated struct{ Statuses bool } ``` Helpers: - `allWorkspaceRelated()` / `allLatestBuildRelated()` construct a fully populated tree. Passed by all current callers so behavior is unchanged. - Zero value (nil subtree) includes nothing but the workspace itself. - `(*latestBuildRelated).appStatuses()` is a nil-safe accessor for the deep `apps.statuses` leaf so call sites descend the tree without a guard chain. Rationale for the pointer tree over a flat struct plus `withAncestors()`: the invariant holds by construction (illegal state is unrepresentable) instead of relying on a normalization step a caller must remember. `workspaceBuildsData` takes the `*latestBuildRelated` subtree directly. ### Query gating `workspaceData(ctx, workspaces, cfg)`: - `GetTemplatesWithFilter` gated on `Template`. - `GetLatestWorkspaceBuildsByWorkspaceIDs` gated on `LatestBuild`; when off, skip `workspaceBuildsData` entirely (empty builds). - `GetLatestWorkspaceAppStatusesByWorkspaceIDs` (workspace-level LatestAppStatus) gated on `AppStatuses`. `workspaceBuildsData(ctx, builds, cfg)` gates each stage: - jobs (`GetProvisionerJobsByIDsWithQueuePosition`) + eligible provisioner daemons gated on `Job`. Only one job query exists and it always computes the queue position, so `QueuePosition` does not gate a separate query yet; it is modeled for hierarchy/Phase 2 completeness. A cheaper job query without the window-function ranking is a worthwhile follow-up but is out of scope for a simple Phase 1 (needs new SQL + sqlc gen). - resources gated on `Resources`; metadata on `Metadata`; agents on `Agents`; apps on `Apps`; app statuses on `AppStatuses`; scripts on `Scripts`; log sources on `LogSources`; template versions on `TemplateVersion`. Downstream `convertWorkspaceBuilds` already tolerates empty slices, so squashing queries yields zero/nil fields without extra work. ## Testing (comprehensive, mocked database.Store) Internal tests (`coderd/wsrelateddata_internal_test.go`) build `&API{Options: &Options{Database: dbmock.NewMockStore(ctrl)}}` and call `workspaceData` directly. gomock is strict, so setting EXPECT only on permitted queries proves the rest are squashed. Cases: include-all (every query runs); zero value (no related queries run); each single node (template only, latest_build only, job, resources, metadata, agents, apps, statuses, scripts, log_sources, template_version, workspace-level app statuses); the empty-resources short-circuit; and the constructors plus the nil-safe `appStatuses()` accessor at every depth. ## Steps 1. Add `coderd/wsrelateddata.go` (selection tree, constructors, `appStatuses`). 2. Thread the selection through `workspaceData` and `workspaceBuildsData`; gate queries by descending the tree. 3. Update the callers to pass `allWorkspaceRelated()` / `allLatestBuildRelated()`. 4. Add unit tests; run `make fmt`, `make lint`, targeted `go test`. ## Resolved decisions 1. Pointer tree over a flat struct + normalizer, so the ancestor rule is a structural invariant. 2. Cheaper queue-position-free job query deferred to a follow-up. </details> --- *Generated by Coder Agents on behalf of @spikecurtis.* |
||
|
McKayla はな
|
676b9bb06b | fix: only create workspace agents for start builds (#28276) | ||
|
Asher
|
7ba2e6e22a |
fix: add distributed protection for external auth refresh (#27479)
Before refreshing, the caller has to put a lease on the row to prevent any other replicas from trying to refresh using the same token. Callers unable to get a lease enter a holding pattern waiting for the results. Also use this same lease query to re-read the link. This more or less functions the same as the re-read we used to have after a refresh failure, except it also allows us to avoid making the failed request in the first place. The downside is that we have to refetch the link every time even if the caller just fetched it, but since we have to make a database call to set the lease anyway... |
||
|
Michael Suchacz
|
24b83dc727 |
feat: add organization picker to MCP server settings pages (#28150)
Adds an organization picker to the MCP server settings pages so
multi-org deployments can manage every organization's MCP servers
through the UI, not just the default organization's. Closes CODAGT-714
(scoped to MCP servers only; model configs stay deployment-wide until
CODAGT-709 lands).
## Stack Context
Top of the CODAGT-711/717/712/806 stack: MCP server configs are
org-scoped with RBAC, audit logging, per-config ACLs, and owner-scoped
user tokens. This layer makes the admin UI org-aware, resolving the
"default org only" limitation documented in the bottom PR.
## Summary
- The MCP servers list and add pages resolve the working organization
from an `org` search param (falling back to the default org), and render
an organization picker built on the existing `OrganizationAutocomplete`.
The picker only appears when more than one organization is visible, so
single-org deployments see no change.
- Selecting an organization rewrites the search param, refetches the
list for that org, and threads through to the add page so creation posts
to the selected organization.
- The update page now loads the server by ID (new organization-scoped
`GET /api/experimental/organizations/{organization}/mcp-servers/{id}`
frontend binding) instead of listing the default org and searching
client-side, so servers from any organization can be edited. Post-save
navigation returns to the owning organization's list via the server's
`organization_id`.
- Removes the now-unused `getDefaultOrganizationId` helper.
- Review hardening: organization display labels are disambiguated with
the unique org name when effective display names collide (picker
options, trigger text, and accessible name). Access is no longer
restricted to deployment admins: the pages, sidebar item, and nav entry
accept org-scoped `mcp_server_config` permissions, the picker lists only
organizations the user can view, content and affordances (add, edit,
delete) are gated on the selected organization's permissions, and
deployment admins keep full access everywhere.
## Testing
Storybook interaction stories cover: picker hidden with one org,
switching orgs refetches and swaps list contents, add posting to the
URL-selected org, add posting after an in-form org switch (asserting
typed form state survives the switch), and update loading by ID.
Red-green verified by breaking the search-param resolution and observing
the new stories fail.
> Mux acted on Mike's behalf for this PR.
|
||
|
david-fraley
|
8203b2ebaf | feat!: hide Coder Tasks behind the enable-ai-tasks flag (#28008) | ||
|
Michael Suchacz
|
a3eb1ee930 |
fix: gate MCP user tokens on owner-scoped personal RBAC (#27945)
Re-gates every `mcp_server_user_tokens` dbauthz method from `rbac.ResourceDeploymentConfig` onto owner-scoped personal actions, mirroring the `external_auth_links` prior art. These are per-user credential rows, not deployment config, and the old gate forced every call site to elevate with `dbauthz.AsSystemRestricted`. ## Summary - Reads gate on `ActionReadPersonal` and writes on `ActionUpdatePersonal` against the token owner (`MCPServerUserToken.RBACObject()` derives from `user_id`). The id-keyed refresh and refresh-failure mutations resolve the owner via a `GetMCPServerUserTokenByID` prefetch inside dbauthz; the mutation queries cannot retarget a row's owner, so there is no fetch-authorize race. - All seven token-path `AsSystemRestricted` elevations in `coderd/mcp.go` and all four in chatd are removed along with their `//nolint:gocritic` markers; handlers use the request actor. - The daemon-wide chatd subject keeps only `ActionReadPersonal` on `ResourceUser`. Background token refresh and failure persistence use a new per-user `AsChatdTokenOwner` subject holding only user-level `ActionUpdatePersonal` scoped to the token owner (both mutations resolve the row via the underlying store, so the subject needs no read grant), and chatd never holds site-wide personal-write access. `TestAsChatd` pins the complete `ResourceUser` action set and `TestMCPServerUserTokensAuth` exercises both subjects against the real authorizer. - New strict-authorizer test (`rbac.NewAuthorizer` against real rows): owner passes get/upsert/delete, a stranger fails, and an org admin of another org fails. No user-visible behavior change. Stacked on #27944. Completes the MCP org-separation stack. Closes https://linear.app/codercom/issue/CODAGT-806 UAT: regression pass on a dogfood instance: OAuth2 connect redirect, member list/status with redaction, self-scoped disconnect (cannot touch another user's token), end-to-end chat MCP tool invocation, and lower-stack audit/ACL smoke checks. > Mux (AI agent) authored this PR on Mike's behalf. <!-- mux-attribution: model=claude-fable-5 thinking=high --> |
||
|
Michael Suchacz
|
7ca7c30f40 |
feat: add group and user ACLs to MCP server configs (#27944)
Adds group and user ACLs to org-scoped MCP server configs so
organizations can restrict specific MCP servers to subsets of members,
mirroring the template ACL pattern.
## Summary
- Migration adds `group_acl`/`user_acl` JSONB columns (nested `{"<id>":
{"permissions": [...]}}` shape) and seeds every existing config with its
organization's Everyone group read entry (the Everyone group ID equals
the org ID), so member access is unchanged by default. Creation seeds
the same entry.
- The blanket org-member read grant from the base PR is replaced by ACL
evaluation: Rego requires org membership for every ACL grant, and
`GetAuthorizedMCPServerConfigs` compiles ACL-aware SQL filters.
- New `ActionShare` (org admins) gates `GET|PATCH
/api/experimental/organizations/{organization}/mcp-servers/{mcpserverconfig}/acl`
(nested under the organization like the rest of the config surface).
PATCH validates principals against the config's organization, merges
sparse updates under a row lock, stamps `updated_by`/`updated_at`, and
is audited as a Write with `Old` captured before authorization. A config
deleted concurrently between the middleware fetch and the locked
re-fetch is concealed as 404, matching the update and delete handlers.
ACL columns are tracked in the audit table.
- ACL management is available in all editions (no enterprise
entitlement), documented in the MCP servers page. Revoking an ACL does
not retro-strip already-selected configs from existing chats; new
selection is blocked at chat create.
- Force On respects the ACL: the forced set is loaded as the chat owner,
so a `force_on` server whose ACL denies the owner never attaches at
create, send, or generation time.
- No rolling-upgrade machinery: upgrades run in scheduled maintenance
downtime, so the migration only backfills existing rows; the API sets
the Everyone read grant explicitly on every insert.
Stacked on #27943. Part of the MCP org-separation stack.
Closes https://linear.app/codercom/issue/CODAGT-712
UAT: verified on a dogfood instance with two members and a custom group:
Everyone-seed default visibility, group grant with Everyone removal
(non-member loses list/fetch/selection), foreign-principal rejection,
non-admin share denial (audited 403 / concealed 404), user_acl restore,
and audited ACL diffs.
> Mux (AI agent) authored this PR on Mike's behalf.
<!-- mux-attribution: model=claude-fable-5 thinking=high -->
|
||
|
Michael Suchacz
|
299e72ad30 |
feat: audit MCP server config changes (#27943)
Adds enterprise audit logging for MCP server config create, update, and
delete, with strict secret redaction. MCP configs hold credentials
(OAuth2 client secrets, API keys, custom headers), so admin changes to
them need an audit trail.
## Summary
- `enterprise/audit/table.go` gains an `MCPServerConfig` entry
enumerating every column: `oauth2_client_secret`, `api_key_value`, and
`custom_headers` are `ActionSecret` (never appear in diffs); dbcrypt
`*_key_id` bookkeeping, IDs, and timestamps are ignored; the remaining
config fields, including the endpoint URL fields, are tracked so
auditors can see which endpoints a config points at.
- Type registration in `coderd/audit` (diff, request, resource target
with org attribution), `codersdk/audit.go`, and a `resource_type` enum
migration.
- Handlers wire `audit.InitRequest`: create records `New`; update and
delete record `Old` from the param middleware before the
write-authorization check, so a readable-but-not-writable caller
produces an audited 403 while read-denied callers stay concealed as
unaudited 404s.
- Tests: create/update/delete audit entries, write-denied and
delete-denied 403 auditing, cross-org concealment producing zero
entries, and a serializer-level regression test proving none of the
three secret classes can reach a serialized diff.
- Review round: MCP config audit entries link to
`/ai/settings/mcp-servers/{id}`, audit table comments are trimmed per
review, and a fault-injection test pins that a config row surviving a
failed post-discovery credential update still gets its creation audit
entry.
Stacked on #27942 (org-scoped MCP configs). Part of the MCP
org-separation stack.
Closes https://linear.app/codercom/issue/CODAGT-717
UAT: verified on a trial-licensed dogfood instance: audit entries for
the full CRUD lifecycle with correct actor/org/target, redacted secrets
in the update and OAuth2 create diffs, and a full plaintext scan of the
audit dump finding zero secret leaks.
> Mux (AI agent) authored this PR on Mike's behalf.
<!-- mux-attribution: model=claude-fable-5 thinking=high -->
|
||
|
Michael Suchacz
|
f7a0de7a11 |
fix: harden org-scoped MCP config chat gating, updates, and visibility (#28065)
Hardens the org-scoped MCP server config surface from #27942 with fixes and regression pins that are independent of the core cutover: - Keep chats sendable after a selected MCP server is disabled (persisted selections are exempt from message-time rejection). - Keep cached MCP selections usable after a background refetch error (gate the composer on missing data, not `isSuccess`). - Merge PATCH updates onto the current row so unset fields are not clobbered. - Give auditors the full management view of MCP configs their audit logs reference (site and org auditor roles). - Regression pins: frozen OAuth2 callback path, cross-org concealment for item routes, OAuth callback token binding, and disconnect responses indistinguishable from nonexistent config IDs. - Storybook: semantic textbox queries in MCP loading stories; docs note that the MCP settings page needs deployment access. ## Stack context Part of the MCP org-separation stack (CODAGT-711 org scope -> apidocs -> hardening -> CODAGT-717 audit -> CODAGT-712 ACLs -> CODAGT-806 token RBAC). Split out of #27942 to keep the core cutover reviewable; each change here builds on the org-scoped routes and chat gating introduced below it. > Mux (AI agent) authored this PR on Mike's behalf. <!-- mux-attribution: model=claude-fable-5 thinking=high --> |
||
|
Michael Suchacz
|
f2bc9ab1f5 |
docs: complete swagger annotations for organization-scoped MCP routes (#28064)
Adds the missing swagger annotations for the eight organization-scoped MCP server config routes introduced in #27942 and checks in the regenerated API artifacts (`coderd/apidoc`, `docs/reference/api`). No behavior changes: 58 hand-written annotation lines, the rest is generated output. ## Stack context Part of the MCP org-separation stack (CODAGT-711 org scope -> apidocs -> hardening -> CODAGT-717 audit -> CODAGT-712 ACLs -> CODAGT-806 token RBAC). Split out of #27942 to keep the core cutover reviewable; these routes live under `/api/experimental`, where main already ships several MCP handlers without annotations, so the base PR is consistent with existing precedent until this lands. Closes nothing on its own; documentation completion for CODAGT-711. > Mux (AI agent) authored this PR on Mike's behalf. <!-- mux-attribution: model=claude-fable-5 thinking=high --> |
||
|
Michael SuchaczandMathias Fredriksson
|
443e3b9b80 |
feat!: org-scope MCP server configs with RBAC (#27942)
Moves `mcp_server_configs` from deployment scope to organization scope
so each organization fully controls the MCP servers its members can use
with Coder Agents.
## Summary
- Migration: adds `organization_id` (NOT NULL, FK) and keeps existing
rows as the default organization's originals with credentials intact.
Other organizations start with no MCP servers and configure their own;
nothing is copied across organizations. Chats outside the default
organization keep any now-cross-organization `mcp_server_ids` entries;
the runtime already ignores IDs that do not resolve in the chat's
organization, so no data rewrite is needed. Slug uniqueness becomes
`(organization_id, slug)`.
- RBAC: new org-scoped `ResourceMCPServerConfig` with regosql converter
and `GetAuthorizedMCPServerConfigs`; org admins get in-org CRUD, org
members get read (replaced by ACL evaluation in the follow-up ACL PR in
this stack).
- API: all config routes nest under the organization, matching
templates: `POST|GET
/api/experimental/organizations/{organization}/mcp-servers` and
`GET|PATCH|DELETE .../mcp-servers/{mcpserverconfig}` (plus
`oauth2/connect`), resolved by a read-only param middleware that
conceals read-denied and cross-organization access as 404. Two routes
stay on the frozen `/api/experimental/mcp/servers/{mcpServer}` block:
the OAuth2 callback (the redirect URI baked into existing AS-side client
registrations) and `oauth2/disconnect`, which must remain reachable by
users removed from the organization so they can still revoke their token
grant.
- Chat runtime: selection validation and generation resolve configs
strictly by IDs, enabled state, and the chat's organization in SQL;
requested duplicates are normalized; invalid or cross-org IDs are
rejected with the precise ID list. IDs already persisted on a chat are
exempt from message-time rejection so disabling a selected server never
blocks sends; generation skips servers that are no longer usable.
- Frontend: API layer and admin settings pages target the new endpoints.
The admin page manages the default organization's servers; the org
picker is tracked separately (CODAGT-714).
- Security hardening from review: OAuth user grants are additionally
bound to `oauth2_revocation_url` (changing it invalidates grants, and a
racing OAuth callback gets 409 instead of recreating a grant).
Stack-wide SSRF protection for MCP config-directed traffic was split
into its own PR at the top of this stack (#28242) to keep this diff
reviewable; this PR keeps main's existing discovery IP-range guard.
OAuth2 auto-discovery now completes before the config row is inserted: a
failed discovery persists nothing, and there is no provisional row that
concurrent updates could race against.
Two follow-up PRs in this stack were split out to keep this diff
reviewable: #28064 completes the swagger annotations for the moved
routes (main already ships these experimental MCP handlers unannotated),
and #28065 carries hardening fixes and regression pins on top of the
cutover.
- Force On enforcement (landed on main mid-review) is org-scoped: the
forced set is read per chat organization
(`GetForcedMCPServerConfigsByOrganization`), so another organization's
`force_on` server never attaches to a chat.
## Breaking changes (experimental API)
The MCP server config endpoints move from the deployment-scoped
`/api/experimental/mcp/servers` block to organization-nested paths:
`POST|GET /api/experimental/organizations/{organization}/mcp-servers`
and `GET|PATCH|DELETE
/api/experimental/organizations/{organization}/mcp-servers/{mcpserverconfig}`
(plus `oauth2/connect`). The old paths are removed, so API consumers
must supply an organization. Two routes intentionally stay on the frozen
`/api/experimental/mcp/servers/{mcpServer}` block: the OAuth2 callback
(its redirect URI is baked into existing AS-side client registrations)
and `oauth2/disconnect` (must remain reachable by users removed from the
organization). These endpoints are under `/api/experimental`, so no
deprecation window is provided.
## Rolling upgrades
During a rolling deploy, an old replica creating an MCP config can fail
the new `NOT NULL organization_id` constraint until it is upgraded
(reads are unaffected: old binaries' generated queries select their own
column lists). This matches the repo's existing precedent for additive
NOT NULL migrations (000562) and affects only the admin config-create
path in the upgrade window.
Upgrades are expected to run in scheduled maintenance downtime with the
database locked during migration, so the migration ships no
rolling-upgrade compatibility machinery. The down migration deletes
organization-created configs (their chat references are cleaned by the
000510 delete trigger) and restores deployment-wide slug uniqueness.
Part of the MCP org-separation stack (CODAGT-711 -> CODAGT-717 audit ->
CODAGT-712 ACLs -> CODAGT-806 token RBAC).
Closes https://linear.app/codercom/issue/CODAGT-711
UAT: validated end to end on a two-org dogfood deployment, including a
real pre-migration to post-migration upgrade, cross-org isolation
(404s), same-slug-two-orgs, chat selection gating, and a live MCP tool
call through the org-scoped generation path. The migration was later
revised to keep existing rows in the default organization only (no
per-organization copies); that revision is covered by the migration test
suite.
> Mux (AI agent) authored this PR on Mike's behalf.
<!-- mux-attribution: model=claude-fable-5 thinking=high -->
---------
Co-authored-by: Mathias Fredriksson <mafredri@gmail.com>
|
||
|
Hugo Dutka
|
c2bb62446d |
fix(coderd): remove per-chat system prompt limit (#28294)
Per-chat system prompts longer than 10,000 characters are rejected during chat creation. Remove that validation so chat creation accepts longer system prompts within the existing request-body limit. |
||
|
Ben Potter
|
8a7e8d9d5b |
refactor: move chat prompt sanitization into codersdk (#28283)
Exports `SanitizePromptText` (and its helpers) from `codersdk` instead of `coderd/x/chatd`, so API consumers can sanitize prompt text exactly the way the server stores it. ## Why coder/terraform-provider-coderd#412 adds a `coderd_chat_system_prompt` resource whose `system_prompt` attribute compares values by their sanitized forms (otherwise the trailing newline from `file("system-prompt.md")` is perpetual drift, since the server stores the sanitized value). That currently requires a mirrored copy of the sanitizer in the provider, which can silently rot. Per review there ([discussion](https://github.com/coder/terraform-provider-coderd/pull/412#discussion_r3801110795)), the sanitizer should be exported from `codersdk` and imported instead. ## What - `coderd/x/chatd/sanitize.go` → `codersdk/promptsanitize.go` (pure move; package + doc-comment note about why it lives in codersdk) - `coderd/x/chatd/sanitize_test.go` → `codersdk/promptsanitize_test.go` - Callsites updated (`exp_chats.go`, `chatd.go`, `subagent.go`, `context_prompt.go`); no wrapper left behind, single source of truth - No behavior change > Generated by Coder Agents on behalf of @bpmct |
||
|
Paweł Banaszewski
|
63641b98c8 |
fix: treat a missing serve endpoint as a fatal dial error (#27864)
Adds 404 as a terminal error for establishing DRPC connection. A standalone AI Gateway pointed at a coderd that does not expose `/api/v2/ai-gateway/serve` gets a 404, which the connect loop classified as transient and retried forever. Redialing cannot fix a missing endpoint. 404 now is treated as terminal handshake failure. `--url` is expected to point directly at coderd, so a 404 from an intermediary is not distinguished. Refs https://linear.app/codercom/issue/AIGOV-320/write-connection-tests --- Generated with Coder Agents. |
||
|
Bobby Ho
|
663f41ffa9 |
feat: derive OAuth2 client type from token_endpoint_auth_method (#28043)
Adds an OAuth2 client type (public vs confidential, RFC 7591 §2) derived from the requested auth method instead of hardcoded confidential. The type is stored and guarded here, but no endpoint enforces on it yet; public behavior at the token endpoint follows in the next PR in the stack. - Client type is derived once and reused by both registration and redirect URI validation, so they can't disagree - IsPublic() fails closed: an unrecognized or missing value reads as confidential - RFC 7592 update (PUT) now rejects moving a client between public and confidential (400) instead of silently flipping it when the auth method is omitted - Discovery still doesn't advertise "none"; follows once the token endpoint honors it ### Behavior by client shape `client_type` is derived from `token_endpoint_auth_method` at POST and pinned at PUT. RFC 7592 GET/PUT authenticate with the registration access token, not the client secret, so neither endpoint reads a secret. | Registered with | Stored `client_type` / method | GET reports | PUT that flips the method | |------------------------------------|---------------------------------------|-----------------------|-----------------------------------------| | omitted, or `client_secret_basic` | `confidential` / `client_secret_basic` | `client_secret_basic` | `none` → 400 `invalid_client_metadata` | | `none` (new) | `public` / `none` | `none` | `client_secret_*` → 400 `invalid_client_metadata` | | `none` (before this PR) | `confidential` / `none` | `none` | either → 200, type stays `confidential` | - PUT still replaces every other RFC 7591 field. `client_type` is the only pinned one; the method may move within a type (`client_secret_basic` ↔ `client_secret_post`). - Row 3 is the only shape where the two columns disagree. The guard fires only on a method change that crosses the type line, so those clients keep managing themselves instead of being locked out of their own configuration endpoint. - The token endpoint does not consult `client_type` yet, so every client still authenticates with a secret and registration still issues one. Split out of #27873, second in the stack (on top of #28041). Refs https://linear.app/codercom/issue/ENG-3029/oauth2-support-public-client |
||
|
Andrew Aquino
|
71e95a3611 |
feat(coderd/tracing): correlate request logs and spans by client_session_id (#27671)
## What Adds `client_session_id` correlation to coderd's HTTP request handling, per the [Connection log collection and correlation RFC](https://www.notion.so/coderhq/Connection-log-collection-and-correlation-36ed579be5928025a56cd11fe58661fb). Clients attach a per-session correlation ID to every API request via W3C baggage using the `client_session_id` key. This change makes coderd's tracing middleware read that baggage member and: - add `client_session_id` to the **per-request log context** so all logs for a request (and the handlers it calls) can be correlated by a single ID, and - set `client_session_id` as a **span attribute** when tracing is enabled. Per RFC requirement 6.1, the value is added to the log context **even when tracing is disabled** (the middleware previously returned early when no tracer provider was configured, so baggage was never read). The `client_session_id` is validated as a 32-character hexadecimal string (a 16-byte value, per RFC requirement 1) to guard against logging arbitrary client-controlled baggage values. ## Scope This is `DEVEX-659` and is intentionally limited to the coderd tracing middleware. It is the first piece of a stack: the web terminal client change (`DEVEX-663`) that generates and sends the `client_session_id` will be stacked on top of this PR. No client currently sends `client_session_id` baggage, so this change is a no-op until the client work lands. ## Testing - `coderd/tracing`: new unit tests cover `validSessionID`, baggage extraction (`sessionIDFromHeaders`), and the middleware end to end, asserting `client_session_id` lands on the log context with tracing enabled **and** disabled, is exposed as a span attribute when tracing is enabled, and that absent/malformed baggage is ignored. - Existing `Test_Middleware` route-matching behavior is unchanged. <details> <summary>Design notes / decision log</summary> - **Where the value is read:** the existing `tracing.Middleware` runs high in the coderd middleware stack (`coderd/coderd.go`), before request-id and request-logger middleware, and already matches the `/api`, `/api/**`, app proxy, and external-auth routes. Reading baggage here means the `client_session_id` is on the context before the request logger and handlers run, so it flows into all downstream `slog` calls that use the request context. This mirrors the existing `request_id` pattern in `httpmw.AttachRequestID` (`slog.With(ctx, ...)` + span attribute). - **Works when tracing is off:** the middleware now gates only on the route matcher, extracts baggage and adds `client_session_id` to the log context for all matched routes, and only then branches on whether a tracer is configured. When a tracer is present, `client_session_id` is additionally set as a span attribute. - **Explicit baggage propagator:** extraction uses `propagation.Baggage{}` directly rather than the global text map propagator, so it does not depend on the global propagator being configured (also makes it deterministic in tests). - **Validation:** only a 32-char hex string is accepted (lower or upper case). Malformed values are dropped rather than logged, preventing log/attribute pollution from arbitrary client-supplied baggage. - **Out of scope for this PR (tracked elsewhere):** client generation/sending of `client_session_id` (`DEVEX-663`, web terminal), the equivalent agent-side middleware (RFC 6.2), `connection_logs.client_session_id` (RFC 12), and additional connection state-change logging (RFC 7-13). </details> --- _Opened by Coder Agents on behalf of @aqandrew._ |
||
|
Jon Ayers
|
821d91fabd | fix: log tailnet tunnel authorization decisions (#27819) | ||
|
Bobby Ho
|
166d92ba73 |
fix: bound request body size on JSON API endpoints (#28168)
## Summary `httpapi.Read` decoded request bodies with no size limit, so a single request could allocate memory without bound. This adds a 4 MiB default ceiling, leaves the endpoints that legitimately need more explicitly exempted, and counts the rejections so a limit set too tight is visible. This is the first of three PRs split out of #28048, covering the endpoints that answer in `codersdk.Response` shape. The OAuth2 decode paths (RFC 6749, RFC 7591) and the SCIM ones (RFC 7644) answer in their own error shapes and follow in separate PRs, along with the lint rule that pins the invariant. Closes PLAT-463. Remediates SEC-416 (CWE-770, CVSS 7.5) and SEC-392. ## Problem `httpapi.Read` calls `json.NewDecoder(r.Body).Decode(value)` with no ceiling, and no middleware in the chain bounds body size. The exposure is pre-authentication: login, OTP, and first-user creation all read a body before any authorization decision is reached. The existing rate limiter bounds request *rate*, which is orthogonal to the memory a single admitted request may consume. ## Fix `Read` is split into `Read` and `ReadLimit`. `ReadLimit` wraps `r.Body` in an `http.MaxBytesReader` and keeps the existing decode and validate logic; `Read` delegates to it with a new `DefaultMaxRequestBodyBytes` of 4 MiB, which covers the 124 remaining non-test callers at a single site. `http.MaxBytesReader` composes as tightest-wins, so the handlers that pre-wrapped their own bodies pass their limit to `ReadLimit` rather than wrapping, and each keeps its previous ceiling byte for byte. That matters most for the bulk secrets import at `8 * MaxSecretsFileBytes`: an unconditional wrap inside `Read` would have silently halved it to the default. `TestImportUserSecretsBodyLargerThanDefaultLimit` is the regression guard for that specific failure, and `TestMaxBytesReaderNesting` pins the composition behavior the whole requirement rests on. Every rejection site calls `httpapi.RecordRequestBodyLimit`, which names the limit that tripped on the request's existing log line and marks the request so `coderd_api_requests_too_large_total{reason="request_body"}` counts body rejections apart from the 413s coderd answers for other causes, such as agent log storage overflow. A limit set too tight for a legitimate payload therefore surfaces without waiting for a user report. The limit is a constant rather than a deployment option: an operator raising it to unblock something would reopen the vulnerability as configuration, where a security scan will not find it. A legitimate 413 is answered with a targeted `ReadLimit` on that endpoint. ## Behavior change `POST /api/v2/files` now answers 413 rather than 400 when a request body exceeds `HTTPFileMaxBytes`. It installed that bound already but reported the rejection as a read failure, which leaked the stdlib `http: request body too large` string through `Detail` and kept the largest limit in the tree off the metric. The separate 413 for an oversized expanded archive is unchanged. The task log snapshot endpoint now answers 413 rather than 400 when its 64 KiB cap is exceeded. Routing it through `ReadLimit` also changes its decode-failure message from "Failed to decode request payload." to "Request body must be valid JSON.", which is what every other endpoint answers. Its tests are updated to match both. `coderd_api_requests_too_large_total` is new, so there is no existing query to migrate. It counts the 413s coderd answers, labeled `method`, `path`, and `reason`. `reason="request_body"` is a rejection by one of the limits above; `reason="other"` is a 413 that has nothing to do with body size, such as agent log storage overflow. ## Reading this The commits are ordered to be read in sequence. Commits 1 and 2 are the security fix; commits 3 to 5 are the observability consequences, and commit 3 is the one that touches dashboards. Commit 7 documents the limit on the REST API reference index. Commits 6 and 8 add and revert an exhaustive `@Failure 413` annotation pass, which buried the fix under its regenerated swagger, and cancel out. |
||
|
Mathias FredrikssonandMichael Suchacz
|
d3f08b1983 |
feat: audit chat system instructions changes (#27668)
Adds an audit record for administrative events on the deployment-wide
chat instruction settings (system prompt, the include-default toggle,
and the plan-mode instructions), per CODAGT-719 and operator decision
D5. Each endpoint records under a stable identity: resource type
`chat_instruction_settings`, a fixed resource ID and a human-readable
target ("System prompt", "Plan mode instructions"), so two changes to
one setting share an ID and history-by-setting works. A real change
exports a Write entry with the old-to-new text visible; a
value-identical PUT still upserts and still returns 204 but records
nothing.
Attempts are recorded, not only transitions. Identity is assigned before
the authorization check, so a denied PUT exports a 403 row with an empty
diff (no request content reaches it), a validation failure exports a 400
row, and a write failure exports a 500 row, each with an empty diff; an
operator can tell "nothing changed" from "something changed and capture
degraded" by the status code.
The write path stays authoritative. The advisory lock and, on plan-mode,
the transaction exist only to serve change-detection; if any of that
machinery fails (lock, begin, commit, rollback), the handler runs main's
idempotent write path directly and derives the response from it, so a
member-visible failure of audit-only infrastructure can never replace
main's successful response. Accepted consequence: when the lock cannot
be taken, two concurrent identical writes can produce two rows instead
of one. That is audit degradation, which is allowed; changing a member's
response is not. Write failures keep the exact response the endpoint
produced before this wiring (transaction error for the system prompt,
which was always transactional; the raw write error for plan mode, which
was not), and the full transaction error is logged so rollback failures
cannot vanish.
<details>
<summary>CODAGT-66 plan entry: S1 (verbatim)</summary>
**S1 `feat: audit chat system instructions changes`** (CODAGT-719; base:
main)
- Struct: `database.ChatSystemPromptSettings{ID uuid.UUID; SystemPrompt
string; IncludeDefaultSystemPrompt bool; PlanModeInstructions string}`
in `coderd/database/types.go` (ticket-sketched shape; one struct, both
endpoints).
- Registration: union entry (diff.go), table.go entry (`id`
ActionIgnore, other three ActionTrack), `AuditActionMap` Write-only;
four request.go cases (`ResourceTarget` "", `ResourceID` from struct,
`ResourceType` new enum value `chat_system_prompt_settings`,
`ResourceRequiresOrgID` false with the "Artificial ID / deployment
singleton" comment convention).
- Migration: `ALTER TYPE resource_type ADD VALUE IF NOT EXISTS
'chat_system_prompt_settings';` comment-only no-op down (000558 shape);
number picked at push per the numbering constraint.
- codersdk: constant + prose `FriendlyString` ("chat system prompt
settings"); `TestAuditDBEnumsCovered` forces both. `coderd/audit.go`
presentation switches: rely on safe defaults (no link, generic
description); no FE changes (filter label falls back to capitalized
value; acceptable per precedent).
- Wiring `putChatSystemPrompt` and `putChatPlanModeInstructions`:
InitRequest with Action Write; artificial `ID: uuid.New()` on `New` only
when a change is detected; no-op suppression by leaving both aReq sides
unset (nil resource IDs skip the log, request.go skip rule); the write
path itself stays byte-identical (upserts still run unconditionally).
- `putChatSystemPrompt` (writes two keys conditionally in one existing
tx): inside that tx, read the pair via `GetChatSystemPromptConfig` for
`Old`, perform the conditional writes exactly as today, then RE-READ the
pair for `New`. The re-read is load-bearing:
`include_default_system_prompt` is computed from the toggle row AND the
prompt, so a prompt-only write can flip the effective value without the
request carrying the pointer. `PlanModeInstructions` stays zero on both
sides.
- `putChatPlanModeInstructions` (no tx exists today): wrap its
read-upsert in `InTx` (behavior-preserving: same single write);
`Old`/`New` populate only `PlanModeInstructions`; the two system-prompt
fields stay zero on both sides; no cross-key reads.
- Change detection compares the populated payload fields only (never the
artificial ID).
- Tests: handler-level coderdtest with `audit.NewMock()` asserting Write
entry on change and NO entry on a value-identical PUT, for both
endpoints (this also exercises `ResourceRequiresOrgID` end to end); the
fallback-flip case (no explicit include-default row, nonempty prompt set
to empty, effective boolean flips: entry emitted with the boolean diff);
diff assertions (old->new prompt text tracked, not secret) in
`enterprise/audit/diff_internal_test.go`; `TestAuditableResources`
passes by construction.
- Bookkeeping at PR open: correct CODAGT-719's no-op premise ("matches
the existing 204-on-unchanged behavior" does not exist on main;
suppression is new, write path unchanged).
- Review focus: Old capture and the New re-read inside the tx (three of
four existing singletons never set Old; do not copy them; and the
computed include-default value makes a naive New construction wrong);
the skip-on-no-op mechanism; prompt text deliberately visible in diffs.
</details>
Note: the plan excerpt above predates operator decision D5 (2026-07-30),
which this PR implements: the resource type is
`chat_instruction_settings` (not `chat_system_prompt_settings`), each
setting carries a stable ID and a display-name target (not a per-write
artificial ID and an empty target), no-op suppression runs through
`InitRequestWithCancel` (not the nil-ID skip), and attempts (denied,
failed, capture-degraded) record rows with real statuses and empty
diffs. Ticket bookkeeping for CODAGT-719 was corrected on Linear at
kickoff: the ticket's "matches the existing 204-on-unchanged behavior"
premise does not exist on main; suppression is new, and the write path
is unchanged.
> 🤖 This PR was created with the help of Coder Agents, and _will be_
reviewed by a human. 🏂🏻
---------
Co-authored-by: Michael Suchacz <203725896+ibetitsmike@users.noreply.github.com>
|
||
|
Michael Suchacz
|
affeeaf9c8 |
feat: extend agent chat MCP tools for remote UAT evidence loops (#28233)
Extends the Agent-chat MCP tools so an unattended UAT evidence loop can
fetch artifacts, monitor long runs, and find prior runs without burning
model context.
## Backend
- New `chat_files_token` crypto key feature (migration 000571) with
rotator support and a dedicated signing keycache on coderd.
- `POST /api/experimental/chats/files/{file}/download-url`
(authenticated) mints a short-lived (5 min) signed URL and returns it
with `sha256`, `size_bytes`, `name`, `mime_type`, and `expires_at`.
- `GET /api/experimental/chats/files/{file}/download?token=` (no session
token) redeems the signed URL: verifies the JWS, requires the token's
`file_id` to match the path, and re-checks the minting user's RBAC
access live at redemption. Clients can `curl -o` artifacts with zero
credentials in the URL consumer.
- `ChatFileMetadata` gains `size_bytes` (via `octet_length`, no bytes
fetched).
## MCP tools (`codersdk/toolsdk`)
- `coder_download_chat_file`: by `file_id` or `chat_id`+`file_name`;
returns the signed URL plus checksum and size instead of base64.
- `coder_await_chat`: blocks (bounded `wait_secs`, 1-120) until a chat
leaves `running`/`interrupting`, using the existing watch stream with
subscribe-before-read.
- `coder_list_chats`: label, query, and limit filtering; chat
projections now include labels.
- `coder_get_chat_messages`: `after_id` forward cursor with
`next_after_id` (exact incremental reads), plus per-message `files`
metadata so artifact-bearing messages are identifiable.
- `coder_get_chat`: file listings now include `size_bytes` and
`created_at`.
- `coder_list_templates`: exposes `agents_allowed` for pre-flight
checks.
## Testing
- coderd: mint/redeem happy path with an unauthenticated client,
expired/tampered/file-mismatched tokens, auth still required on the
plain file endpoint, non-owner mint rejection.
- toolsdk: harness + integration coverage for all new/changed tools,
including signed-URL redemption with checksum verification,
forward-cursor exactness, await transition/timeout paths, and label
filtering.
- Remote dogfood UAT (dev.coder.com Coder Agent) passed all six
acceptance scenarios end to end over both MCP transports.
Note: `go test ./codersdk/toolsdk/` has a pre-existing goleak flake on
main (leaked `agentssh` non-PTY session goroutines from SSH exec tests;
reproduced 3/3 on clean `b4971bc49f1`). It is unrelated to this diff.
> Mux acted on Mike's behalf to create this PR.
<!-- mux-attribution: model=claude-sonnet-4-6 thinking=high -->
|
||
|
Michael Suchacz
|
7724ee281a |
feat: defer MCP tool schemas behind a find_tools search (#28225)
## Summary When the `mcp-tool-search` experiment is enabled, chatd stops inlining connected MCP tool schemas into every generation. It instead exposes a built-in `find_tools` tool whose description carries a compact catalog of the deferred tools, and only ships full JSON schemas for tools the model has activated by searching or by calling them directly. Closes [CODAGT-760](https://linear.app/coder/issue/CODAGT-760). ## Problem Tool-heavy agent configurations (GitHub, Linear, Notion, and dev-tooling MCP servers) inline over 100k tokens of tool schema definitions into every generation. Initial uncached requests reached ~216k tokens with time-to-first-token close to nine minutes, while the model typically invokes only a handful of tools per turn. ## How it works - `decideMCPToolSearch` defers external and workspace `.mcp.json` MCP tools whenever the experiment is enabled. Native, dynamic, provider, skill, and transport tools are never deferred. - `find_tools` embeds a server-grouped catalog in its tool description (degrading to names-only, then counts-only, then a constant-size summary past a context-scaled size cap) and scores keyword matches across tool names, descriptions, parameter schemas, and server metadata. Queries can scope to one server with a `server:` prefix, and exact `names` arguments always activate. - Activation state is ephemeral: it is re-derived each generation from surviving chat history (`find_tools` results and direct calls to deferred tools), so activations naturally lapse when compaction summarizes them away. Aggregate activated schema weight is capped at 10% of the context window, shedding the least recently activated schemas first; `find_tools` shares that budget across parallel calls in one step. No new persistence. - Deferred tools stay registered for execution, so the model can call a cataloged tool directly without searching first; the schema is activated for subsequent steps. - Fail-open: the experiment being disabled, an empty candidate set, or an MCP tool named `find_tools` all disable deferral, leaving today's behavior byte-identical on the wire. - Prometheus counters/histograms track `find_tools` calls, matches, activations, and deferred token weight. - The conversation timeline renders `find_tools` calls with a collapsed search summary and expandable match list, falling back to the generic renderer on malformed payloads. ## Validation - Unit tests for the catalog, matcher, experiment-gated decision, and activation derivation; end-to-end chatd generation tests covering search-then-call, direct-call activation, experiment-off wire parity, compaction lapse, and subagent tool gating. - Storybook interaction tests for the timeline rendering and malformed-payload fallback. - Remote dogfood UAT on dev.coder.com passed: deferral with a real MCP server and Anthropic model, direct calls without prior search, activation persistence across turns, experiment-off parity, and clean UI/console. > Disclosure: Mux (AI agent) authored this PR on Mike's behalf. |
||
|
J. Scott Miller
|
62f4afbb60 |
perf(coderd): build the workspace build fan-out maps once per batch (#28074)
`convertWorkspaceBuild` rebuilt seven maps from the caller's global slices on every call and rescanned the provisioner daemon rows to filter by job ID. `convertWorkspaceBuilds` calls it once per build with identical slices, so map construction cost `O(builds x rows)` where `O(rows)` suffices — quadratic in the number of workspaces on `GET /api/v2/workspaces`. The maps move into a `workspaceBuildIndex` built once per batch, keyed exactly as before and now including daemons by job ID. `convertWorkspaceBuild` takes the index instead of eight slices. Its parent already hoisted `workspaceByID`, `jobByID`, and `templateVersionByID` out of the same loop; this makes the rest consistent. Agents are sorted while the index is built, so a resource read by several builds is sorted once rather than once per build. Same comparator over the same rows, so the order is unchanged; `TestConvertWorkspaceBuildsAgentOrder` covers it. `BenchmarkConvertWorkspaceBuilds`, 5 resources x 2 agents x 4 apps per build: | builds | ns/op | B/op | allocs/op | | --- | --- | --- | --- | | 1 | 48.3k -> 48.1k | 121k -> 125k | 342 -> 359 | | 25 | 12.7M -> 1.34M | 38.0MB -> 3.2MB | 69,621 -> 8,347 | | 100 | 186M -> 5.67M | 596MB -> 13.0MB | 1,024,941 -> 33,080 | Single-build conversion is a wash (one extra struct allocation); the quadratic term is gone. Addresses the map-allocation half of PLAT-386 / #27205. Bounding the page size is separate (#28040) and does not remove this cost: at 100 workspaces per page it is still 100 passes over every resource, agent, app, script, log source, status, and daemon row in the page. --- Created with Coder Agents on behalf of @jscottmiller. |
||
|
Michael Suchacz
|
119f2b1dd9 |
feat: limit concurrent chat agents with pooled admission (#27902)
Limits concurrent chat generation on capped deployments to 5 root chats and 10 delegated subagent chats. The pools are deployment-wide and independent, so delegated work can continue while root capacity is full. The default caps live in AGPL code. Enterprise contributes only a licensing unlock, so unlicensed deployments stay capped and cannot fail open. Licensed deployments are uncapped while Agent Hours usage stays below an explicit hard limit. Deployments without a hard limit remain uncapped, and reaching the Agent Hours allocation only triggers warnings. Admission happens before a worker takes chat ownership. Capped deployments serialize admission across replicas with a transaction-scoped advisory lock and derive active and queued state from current ownership plus fresh runner heartbeats, rather than persisted queue markers or per-replica state. The acquisition query returns a bounded, pool-interleaved candidate set instead of ranking the whole backlog; a migration replaces the acquisition index with a pool-aware one. Refused chats stay running but unowned, and interrupt requests bypass admission so users can stop queued or over-cap chats. The single-chat API derives `queued_for_capacity` from live pool state; list endpoints do not report it. The UI polls that value every 5 seconds while a chat is running and shows a callout when the chat is waiting for capacity. Updates the administrator documentation and deployment-wide Prometheus gauges for active and queued agents. Replica-level values must be aggregated with `max`, not `sum`. > Mux updated this PR on Mike's behalf. |
||
|
Cian Johnston
|
6079c514ee |
fix: follow-up fixes for conditional VCS requests (#27711)
Follow-ups from #27627 - Memoizes `Config.Git()` with a mutex so the provider's ETag response cache survives across calls. Only successful construction is cached; errors are retried. - Moves the HTTP client onto `Config.HTTPClient`, wired through `ConvertConfig`, so `Git()` no longer takes a per-call argument that would be silently ignored after memoization. - `newGitHub` and `newGitLab` now return `(Provider, error)`, eliminating the typed-nil-interface class in `gitprovider.New` rather than the single instance. - Gates the 304 branch on a `haveCached` flag instead of a nil body check. - Only caches bodies that decode successfully, preventing poisoned entries. - Keys the response cache on the full token digest rather than a truncated prefix. - Tests added: `TestConfigGitMemoizesProvider`, `TestConfigGitRetriesOnConstructorError`, `TestGitLabConstructorErrorReturnsNilInterface`, `TestResponseCacheStore`, `TestConditionalRequestReuse/MalformedResponseNotCached`; `TestConvertYAML/CustomScopesAndEndpoint` now asserts `Config.HTTPClient` wiring. Follow-ups tracked in #28139, #28140, #28141, #28142. > 🤖 Generated by Coder Agents on behalf of @johnstcn. |
||
|
Jaayden Halko
|
fa8ffe4eda |
feat: report agent runtime hours usage in entitlements (#27985)
Populate `FeatureAgentRuntimeHours.Actual` on every entitlements refresh for licenses that grant the feature. A new `GetTotalUsageHBAgentRuntimeV1` query sums `runtime_ms` over the license's usage period, reading `usage_events` directly: `hb_agent_runtime_v1` is exactly one row per hourly bucket deployment-wide with `created_at` at the bucket start, enforced by the unique partial index introduced in #27983. The measurement reuses the shared `measureUsage` policy from #27984 through a new `AgentRuntimeMsFn` closure (usage publisher subject): failures publish the stable `LicenseAgentRuntimeUsageUnavailableErrorText` and log the cause. Usage is floored to whole hours, matching the unit of the `agent_runtime_hours_*` claims, and at most one warning is emitted per refresh: reaching the allocation supersedes the advisory soft limit. The dashboard renders the soft-limit advisory muted without a sales link and treats the runtime usage-unavailable text as a diagnostic. **Precise usage.** `Feature.ActualMs` (JSON `actual_ms`), set only for `agent_runtime_hours`, carries the exact stored milliseconds backing the floored `Actual` so clients can render fractional hours (e.g. `10.3`). It has the same freshness as `Actual`; the whole-hour warning thresholds are unchanged. **Unlimited licenses.** A license minted with the unlimited (`-1`) allocation decodes to an enabled feature with a nil `Limit` (#27984), so the warning write-back now guards the allocation dereference: no thresholds can exist for an unlimited license, so no runtime hours warning is ever emitted, while `Actual` is still measured and published. `Feature.Compare` is unchanged; for usage-period features the issued-at/end dates decide first, so a metered feature outranks an unlimited one only on an exact timestamp tie, an edge pinned by a `TestFeatureComparison` case and documented on `decodeAgentRuntimeHours`. **Grandfathered premium licenses.** Premium licenses without `agent_runtime_hours_*` claims are now granted the feature disabled with a zero limit over the license term, identical to an explicit `allocation: 0`: usage is measured and published for every Premium deployment, and chatd's pooled admission (#27902) caps concurrent agentic chats until a license with a positive allocation is added. The default carries a fixed early `UsagePeriod.IssuedAt` (2026-08-01, the same mechanism as the managed-agents default) so any license actually carrying the claims outranks it in the `AddFeature` merge regardless of the licenses' relative issue dates; the constant must stay earlier than the earliest legitimately issued claim-bearing license. Zero allocations (explicit or grandfathered) emit no deployment-wide warning banner: those deployments are steered by the in-page upgrade CTA and the concurrency cap. Enterprise licenses are unchanged. Part 3 of a 3-PR stack splitting up #27796 (see there for review history). Stack: #27983 → #27984 → this PR. Closes CODAGT-852. |
||
|
Asher
|
b5d18bb9c9 | feat: add redirect URL override for external auth (#28082) | ||
|
Cian Johnston
|
aa80fa3550 | fix(coderd/externalauth): also retry on 503 (#28218) | ||
|
Susana Ferreira
|
95328f1ead |
fix: label unpriced token usage metric by provider name and type (#28210)
## Problem The `provider` label was inconsistent between AI Gateway metrics. Every metric emitted by the gateway labels `provider` with the provider instance name, for example `anthropic-eu`, while `coder_ai_gateway_cost_control_unpriced_token_usage_records_total` used the provider type, for example `anthropic`. The two could not be correlated on `provider`. The metric was also inconsistent with itself: the path where a provider fails to resolve labelled by instance name, and the path where a model has no price labelled by type. The type is still worth exposing, since prices are keyed on `(provider_type, model)` and that is what an operator needs to add a price. ## Changes - Label the metric with `provider` (the instance name, consistent with the other gateway metrics) and add `provider_type` (the configured type the price is keyed on). - Use `unknown` for `provider_type` when the provider does not resolve to a configured type. - Log the unresolved-provider case at `warn` instead of `info`. A missing price is an expected steady state, but a provider that cannot be resolved is not. - Update the metrics docs and the `metricsdocgen` fixture. Closes [AIGOV-574](https://linear.app/codercom/issue/AIGOV-574) > [!NOTE] > Initially generated by Claude Opus 5, modified and reviewed by @ssncferreira |
||
|
Jaayden Halko
|
20c376a575 |
fix: enforce uniqueness and hour alignment for agent runtime usage events (#27983)
The usage generator writes `hb_agent_runtime_v1` rows with `created_at` at the UTC hourly bucket start and exactly one row per bucket, but nothing in the schema enforced either invariant. A duplicate bucket row under a different id would be double-counted by any consumer summing `runtime_ms`, and a misaligned `created_at` would skew which usage period a bucket is attributed to. This replaces the non-unique partial index `idx_usage_events_agent_runtime` (from migration 000561) with a unique index of the same shape and adds an hour-alignment `CHECK` constraint. Both statements validate existing rows: every supported writer has always produced conforming data, so a pre-existing violator is anomalous and failing the migration loudly beats silently rewriting usage rows. `generateBucket` treats a unique violation on the bucket index as another replica having won the race, mirroring the existing `ON CONFLICT (id)` no-op for committed rows. The `coderd/notifications` sync commit and its revert cancel out (the drift they addressed was fixed on main by #27979); the PR's net diff is only the usage-event changes. Part 1 of a 3-PR stack splitting up #27796 (see there for review history). Stack: this PR → #27984 → #27985. |
||
|
Michael Suchacz
|
521c383f6b |
fix: repair stale chat agent bindings after workspace rebuild (#28152)
## Problem When a chat is bound to a workspace, chatd persists `chats.agent_id` pointing at a specific workspace agent, and it only rebinds on the next chat turn. A workspace stop/start creates a new agent with a new ID in the latest build, so the chat page resolves the stale agent ID to `undefined` and the right sidebar silently drops Terminal, Desktop, Browser, apps, and ports even though the workspace is running. The existing read-time enrichment only filled nil agent IDs and skipped stale non-nil ones, so refreshing did not help until the user sent another message. ## Fix - `coderd/exp_chats.go`: single-chat reads now repair agent IDs that no longer resolve in the workspace's latest build, using the same `agentselect.FindChatAgent` selection chatd uses. A repaired binding also carries the latest build's ID so the response never pairs the new agent with the previous build. Bindings that still resolve are preserved, and repair stays best-effort and response-only (no write-on-read). List reads keep the previous nil-fill-only behavior because validating existing bindings would cost a per-workspace authorization lookup per listed chat. - `site/src/pages/AgentsPage/AgentChatPage.tsx`: the workspace watch update handler detects when a running workspace's latest build no longer contains the chat's bound agent and invalidates the chat query once per chat/build/binding key for immediate recovery, and the chat query polls every 30 seconds while the binding remains unresolved so a transiently failed repair retries even when an idle workspace publishes no further watch events. The watch stream replays the current workspace on every (re)connect, so this covers rebuilds that happen while the page is open or disconnected; page loads are covered by the server-side repair. The workspace-watcher bailout now also keys on `latest_build.id` so a rebuild propagates while the page is open. - `site/src/api/queries/chats.ts`: chat watch events replay the persisted (pre-repair) binding, so the summary merge adopts a snapshot's `build_id` only when the snapshot agrees on `agent_id`, keeping the repaired agent/build pair atomic in the caches. ## Testing - `go test ./coderd -run TestEnrichChatAgentIDs` covering repair, keep-valid, selection-error, list-mode-skips-bound, and no-workspaces cases. - Storybook interaction story `RecoversSidebarAfterWorkspaceRebuild` exercising the watch-event to chat-refetch to sidebar-recovery flow (verified red without the invalidation, green with it). - `pnpm test AgentChatPage.test.ts` covering the binding-resolution predicate. > Mux created this PR on Mike's behalf. |
||
|
Wyatt FryandEthan Dickson
|
a005e5cd22 |
feat: add username and email user search filters (#27922)
## Summary User search can now resolve exact `email:` and `username:` terms through `GET /api/v2/users` instead of only supporting fuzzy free-text matches. The database query already had exact email and username filters; this wires the public search parser and API handler to those filters so clients can ask for a single user by email without fetching every user or depending on substring matching. This is the API half of coder/terraform-provider-coderd#403: that provider PR adds `data.coderd_user.email`, and this PR gives it an efficient exact lookup path. ## Testing - `go test ./coderd/searchquery -run '^TestSearchUsers$' -count=1` - `go test ./coderd -run '^TestGetUsersFilter$' -count=1` - Live API test: - Built local enterprise Coder from this branch. - Started Coder on `http://127.0.0.1:39991` against a clean Postgres database. - Created `lookup-target@example.com`. - Verified `GET /api/v2/users?q=email:LOOKUP-TARGET@EXAMPLE.COM&limit=2` returned exactly one user: ```json { "count": 1, "users": [ { "id": "efc6f909-ce0a-4731-bd2f-6e4df417aaa7", "username": "lookup-target", "email": "lookup-target@example.com" } ] } ``` ---   --------- Co-authored-by: Ethan Dickson <ethanndickson@gmail.com> |
||
|
Bobby Ho
|
1aa3553b52 |
fix(coderd): set Cache-Control: no-store on OAuth2 responses (#28143)
No response from the `/oauth2` route tree set `Cache-Control` at all, so
an intermediary cache or customer-operated reverse proxy was free to
apply a heuristic freshness lifetime to a response carrying a live
credential. RFC 6749 §5.1 and OAuth 2.1 §3.2.3 both make an affirmative
`no-store` directive a MUST for the authorization server.
Adds `httpmw.NoStore`, mounted on the `/oauth2` and
`/api/v2/oauth2-provider` trees, setting `Cache-Control: no-store` and
`Pragma: no-cache` on every response from them. OAuth 2.1 drops `Pragma`
because RFC 9111 §5.4 deprecates it as a request-only field, so sending
both is conformant under either reading. Not operator-configurable,
since both specs say MUST.
## Scope
- **Both trees, not just `POST /oauth2/tokens`.** The mount is one line
either way, and the wider scope also covers DCR registration, client
configuration read and update, the authorize 302 whose `Location` query
carries the code, and `POST /oauth2-provider/apps/{app}/secrets`, which
returns a plaintext client secret. A route added later inherits the
headers, which matters for PLAT-449.
- **A middleware, not a hook in `httpapi.Write`.** Three write paths
never call it: `POST /oauth2/revoke` and `DELETE
/oauth2/clients/{client_id}` write a bare status, and
`writeOAuth2RegistrationError` encodes its own JSON.
- **`/.well-known/*` deliberately excluded.** Public discovery metadata,
and RFC 9728 §5 asks for the opposite treatment. Assertions pin the
exclusion so a later hoist onto a higher router fails CI.
- **Session-credential routes left alone.** `/users/login`,
`/users/otp/change-password`, and `/users/{user}/keys/*` have the same
gap, but PLAT-448 is scoped to OAuth2 and reaching into session auth
changes the risk profile.
Every credential-returning route here is a `POST`, and RFC 9111 §3 bars
heuristic caching of `POST` responses, so this is defense-in-depth
against a non-conformant intermediary rather than a live caching bug.
Both specs say MUST regardless of what caches would actually do.
## Note for PLAT-498
`DELETE /oauth2/tokens` now carries `no-store` and is wrapped in
`apiKeyMiddleware`, which is mounted inside the `/oauth2` tree and
therefore runs after this middleware. It is the one route where both can
write `Cache-Control`, and PLAT-498's write must not replace `no-store`
with something weaker such as `private`. `POST /oauth2/tokens` cannot
overlap, since it deliberately has no `apiKeyMiddleware`.
## Two assumptions testing corrected
- `GET /oauth2/does-not-exist` returns **200**, not 404. Chi runs the
subrouter's middleware chain for unmatched paths, so both headers are
present, but the request falls through to the root router's SPA handler.
The test asserts the headers and deliberately not the status.
- The experiment-disabled case is unreachable from a test binary, since
`RequireExperimentWithDevBypass` short-circuits on `buildinfo.IsDev()`.
A unit test covers the consequence against the `RequireExperiment` it
delegates to.
No schema, `codersdk`, or serpent option changes, so `make gen` produces
no diff. Rollback is a revert.
Refs PLAT-448
|
||
|
Michael Suchacz
|
93c6faf1de |
fix(coderd): send assigned chat model IDs verbatim (#28144)
Fixes #27361 (CODAGT-832). ## Problem When an Agents model was configured under a non-gateway provider type (e.g. Anthropic or OpenAI) with a model ID whose first `/`- or `:`-segment matched a built-in provider name (`anthropic`, `azure`, `bedrock`, `google`, `openai`, `openai-compat`, `openrouter`, `vercel`), `chatprovider.ResolveModelWithProviderHint` parsed it as a canonical `provider/model` reference: the prefix was stripped and the request rerouted to the embedded provider type, overriding the provider the admin explicitly assigned. LLM gateways (e.g. LiteLLM) that namespace their catalogs as `bedrock/...` or `anthropic/...` behind an Anthropic- or OpenAI-type provider failed with an opaque upstream "Model not found", and escaping was impossible (`bedrock/bedrock/...` still rerouted). ## Fix A valid provider hint is now authoritative: `ResolveModelWithProviderHint` returns the assigned provider and the verbatim model ID whenever a hint is present. Canonical `provider/model` and `provider:model` parsing applies only to hint-less resolution paths. Every production call site derives the hint from the model config's explicitly assigned AI provider, so the assignment always wins. The save-time guard rejecting slash-namespaced models on OpenRouter-like providers typed as `openai` (provider named `openrouter` or hosted at `openrouter.ai`) is kept: that combination remains a misconfiguration whose correct fix is the `openrouter` provider type, and rejecting it early beats a confusing upstream error. Its wording no longer claims prefix stripping happens. ## Back-compat note A pre-existing config that relied on stripping (e.g. model `anthropic/claude-x` assigned to an Anthropic-type provider pointing at the real Anthropic API) now sends the prefixed ID verbatim and will get a clear upstream model-not-found error; the admin fixes it by editing the model ID. Nothing in the product ever suggested the canonical form for assigned models. ## Validation - Unit: `TestResolveModelWithProviderHint` updated (hints preserve `bedrock/...`, `anthropic/...`, `provider:...` verbatim; hint-less canonical parsing unchanged), red-green verified against the old ordering. Gateway and openai-type provider routing tests assert verbatim pass-through end to end. - Full `./coderd/x/chatd/...` suites plus `TestCreateChatModelConfig`, `TestUpdateChatModelConfig`, and `TestValidateChatModelConfigProviderModel` pass. - Remote dogfood UAT on real models (PASS): an openai-type provider pointed at a Vercel AI Gateway mount returned a real completion for `anthropic/claude-haiku-4.5`, with trace logs confirming `provider=openai model=anthropic/claude-haiku-4.5` (verbatim, not rerouted); gateway-type (`openai-compat`) routing with `deepseek/deepseek-v4-pro-0813` and the model catalog/picker regressions pass. > Mux acted on Mike's behalf to create this PR. |
||
|
Michael Suchacz
|
d5bb35a49a |
fix(coderd): deflake TestChatMessageWithFiles/FileCapExceeded (#28091)
Fixes the flake tracked in [CODAGT-926](https://linear.app/codercom/issue/CODAGT-926/flake-testchatmessagewithfilesfilecapexceeded). ## Problem `TestChatMessageWithFiles/FileCapExceeded` asserted the rollback of a rejected over-cap send by comparing message counts taken before and after the send. `CreateChat` starts assistant generation asynchronously, so the assistant reply can be persisted between the two reads, making the count check fail even though the rejected message was correctly rolled back ("should have 1 item(s), but has 2"). ## Fix Replace the count comparison with a semantic assertion that the rejected `one too many` message was not persisted, hardened through Codex review rounds: - Scan message history for the rejected marker instead of comparing counts. - Also scan `QueuedMessages`: a busy chat queues the send before file-link validation, so a rollback regression could leave the rejected message queued rather than in history. - Close the queue-promotion race: `getChatMessages` reads history and the queue in two separate database reads, so the assertion first waits for the queue to observe empty; a promoted message must then appear in a fresh history read. ## Verification - Deterministic repro of the exact CI failure signature: waiting for the async assistant reply before the old count assertion reproduced `should have 1 item(s), but has 2` every run. - The new assertion passes under that same forced condition. - Assertion liveness (all temporary red checks reverted): persisting the marker in history fails the history scan; queuing the marker fails the queued scan; queuing the marker and letting it promote fails the post-drain history scan 3/3. - `go test ./coderd -run 'TestChatMessageWithFiles/FileCapExceeded' -count=100` and the full `TestChatMessageWithFiles` parent both pass. > Mux acted on Mike's behalf to create this PR. |
||
|
Michael Suchacz
|
48e1e28638 |
fix(coderd/x/chatd/chattool): make edit_files schema and errors actionable for models (#28121)
## Problem Chat `45b87e40-ffe7-49e5-8932-5fd0bdb9e542` on dev.coder.com failed 57 of 75 `edit_files` tool calls. Every failure was the same: the model omitted `files[].path` (it batched edits per file but only filled in `edits`), and the error relayed back to the model was: ``` POST http://[fd7a:115c:...]:4/api/v0/edit-files: unexpected status code 400: "path" is required ``` The model retried the identical malformed call dozens of times. Two gaps made this sticky: 1. The `edit_files` input schema had no field descriptions, so `path` was only a bare required property. 2. The agent API error reached the model wrapped in HTTP transport noise (method, internal tailnet URL, status code) with no indication of which `files` entry was broken. ## Changes - Add `description` tags to every `edit_files` schema field and state the path requirement in the tool description. - Validate `files` entries in the tool before plan-turn checks and the workspace connection lookup, returning entry-indexed errors such as `files[1].path is required; provide the absolute path of the file to edit; no files in this batch were applied`. - Relay agent API failures with `Message`, `Helper`, `Detail`, and `Validations` from `codersdk.Error` instead of the raw transport-prefixed string. ## Validation - `go test ./coderd/x/chatd/chattool` passes; new tests cover the schema description, entry-indexed validation errors, and transport-noise stripping (each verified red-green by toggling the fix off). - `go build ./...`, `go vet`, and pre-commit (fmt + lint) pass. > Mux created this PR on Mike's behalf. |