## Problem
The `update workspace, new required, mutable parameter added` e2e test
has been flaking consistently
([internal#1328](https://github.com/coder/internal/issues/1328)). The
error:
```
Error: Timed out 5000ms waiting for expect(locator).toHaveValue(expected)
Locator: getByTestId('parameter-field-Sixth parameter').locator('input')
Expected string: "99"
Received string: ""
```
## Root Cause
A race between page navigation and data hydration in `verifyParameters`:
1. The page navigates with `waitUntil: "domcontentloaded"` which does
not wait for API responses to settle
2. React Query may serve stale cached workspace data initially (from
before the update), causing the form to render with empty/old parameter
values
3. The `toHaveValue` assertion uses the default `actionTimeout` of
5000ms which isn't enough time for fresh data to arrive and the form to
re-render
## Fix
- Switch `verifyParameters` navigation to `waitUntil: "networkidle"` to
ensure API responses (workspace data, build parameters) are settled
before the form renders
- Increase the `toHaveValue` timeout to 15s to handle cases where
dynamic parameters hydrate slowly after initial render
Fixescoder/internal#1328
---------
Co-authored-by: blink-so[bot] <211532188+blink-so[bot]@users.noreply.github.com>
## Problem
When switching between chats on the agents page, stream parts could be
lost or applied to the wrong chat due to several race conditions in
`ChatContext.ts`:
1. **`startTransition` deferred parts escape cleanup** —
`startTransition(() => store.applyMessageParts(parts))` defers the state
update. If a chat switch happens between `flushMessageParts` being
called and the transition executing, old-chat parts could apply after
`resetTransientState()` has already cleared stream state for the new
chat.
2. **`message` event has no `chat_id` filter** — Unlike `message_part`,
`queue_update`, and `status` events, the `message` event handler did not
check `streamEvent.chat_id`. While the server-scoped WebSocket makes
this safe in practice, it's an inconsistency in defensive programming.
3. **Brief stale message window on switch** — Between `chatID` changing
and `replaceMessages()` firing (after the query resolves), the store
held old-chat messages while the new WebSocket was already connected.
## Changes
### `ChatContext.ts`
- Added `activeChatIDRef` to track the currently active chat ID
- Guard `startTransition` callback: check `activeChatIDRef` before
applying message parts, discarding them if the chat has switched
- Added `chat_id` filter to `message` event handler, matching the
pattern used by all other event types
- Added `store.replaceMessages([])` to the chatID-change effect so
messages are cleared immediately on switch
### `ChatContext.test.tsx`
Four new tests covering the chat-switch lifecycle:
- WebSocket closure and state reset when chatID changes
- `message` event filtering by `chat_id`
- `startTransition` deferred parts discarded after switch
- Messages cleared immediately before new query resolves
All 13 tests pass (8 existing + 4 new + 1 existing).
## Problem
Non-admin users of the Agents (chat) feature send `model_config_id:
"00000000-0000-0000-0000-000000000000"` (nil UUID) when creating chats,
because the `GET /api/experimental/chats/model-configs` endpoint
requires `policy.ActionRead` on `rbac.ResourceDeploymentConfig`, which
is only granted to admins.
The flow:
1. `AgentsPage.tsx` calls `useQuery(chatModelConfigs())` → hits
`listChatModelConfigs`
2. Non-admin users get a **403 Forbidden** response
3. `chatModelConfigsQuery.data` is `undefined`, so the
`modelConfigIDByModelID` map is empty
4. `handleCreateChat` falls back to `nilUUID` for `model_config_id`
5. The backend rejects the nil UUID: `"Invalid model config ID."`
## Fix
Changed `listChatModelConfigs` to allow all authenticated users to read
model configs:
- **Admin users** continue to see all configs (including disabled ones)
for management via `GetChatModelConfigs`
- **Non-admin users** now see only enabled configs via
`GetEnabledChatModelConfigs` with a system context, which is sufficient
for using the chat feature
This follows the same pattern as `listChatModels`, which already uses
`dbauthz.AsSystemRestricted(ctx)` to allow all authenticated users to
see available models.
Write endpoints (create/update/delete) retain their existing
`ResourceDeploymentConfig` authorization.
## Testing
- Updated `TestListChatModelConfigs/ForbiddenForOrganizationMember` →
`SuccessForOrganizationMember` to verify non-admin users can list
enabled model configs
- All existing chat tests continue to pass
## Problem
When coderd instances are redeployed (e.g. rolling deployment on
dogfood), in-flight chats get stuck in `running` status permanently. The
UI shows them as "thinking" with a spinning indicator, but no worker is
actually processing them. They never error or resume.
## Root Cause
Two bugs combine to cause this:
### Bug 1: Shutdown cleanup uses a canceled context
The `processChat` defer block updates the chat status in the DB when
processing completes. But it uses `ctx`, which `Close()` cancels
*before* the defer runs. The DB transaction silently fails with
`context.Canceled`, leaving the chat in `status=running` with a dead
`worker_id`.
```go
// Close() calls p.cancel() which cancels ctx
// Then the defer tries to use the now-canceled ctx:
defer func() {
err := p.db.InTx(func(tx database.Store) error {
tx.GetChatByIDForUpdate(ctx, chat.ID) // FAILS
tx.UpdateChatStatus(ctx, ...) // FAILS
}, nil)
}()
```
### Bug 2: Stale recovery runs only once at startup
`recoverStaleChats()` was called only once in `start()`, not
periodically. During a rolling deployment, the new instance starts while
the old one is still alive (fresh heartbeat). By the time the old
instance crashes, no one checks again.
## Fix
1. **Use `context.WithoutCancel(ctx)` in the processChat defer** — the
cleanup transaction now completes even during graceful shutdown.
2. **Run `recoverStaleChats` periodically** — a second ticker in the
`start()` loop checks for stale chats at `inFlightChatStaleAfter / 5`
intervals (default: every 1 minute). This catches orphaned chats even
when the instance that owns them crashes without clean shutdown.
## Tests
- `TestRecoverStaleChatsPeriodically` — Verifies chats orphaned *after*
startup are recovered by the periodic loop (not just the startup check).
- `TestNewReplicaRecoversStaleChatFromDeadReplica` — Verifies a new
replica recovers stale chats on startup.
- `TestWaitingChatsAreNotRecoveredAsStale` — Negative test: `waiting`
chats are not incorrectly modified by recovery.
## Problem
The git diff on the `/agents` page had color issues: the editor
background followed light mode but the syntax highlighting used dark
mode (`github-dark-high-contrast`), and the filename header used
light-colored text on a light background.
The root cause was hardcoded dark theme options in the `FileDiff`
component:
```tsx
themeType: "dark",
theme: "github-dark-high-contrast",
```
## Fix
Uses the same theme-aware pattern as every other diff/file viewer in the
codebase (`WriteFileTool`, `EditFilesTool`, `ReadFileTool`, `Tool`,
`response.tsx`):
1. `useTheme()` from `@emotion/react` to read `palette.mode`
2. `getDiffViewerOptions(isDark)` from the shared `utils.ts` module —
returns `github-light` theme for light mode, `github-dark-high-contrast`
for dark mode
3. Reuses `DIFFS_FONT_STYLE` and `diffViewerCSS` constants instead of
inlining duplicates
## Storybook coverage
Added four new stories with real unified diff content:
- **WithDiffDark** — dark mode with a PR link
- **WithDiffLight** — light mode with a PR link
- **NoPullRequestDark** — dark mode, "Files Changed" header
- **NoPullRequestLight** — light mode, "Files Changed" header
The existing stories only covered empty and parse-error states with no
rendered diff.
## Summary
Adds a new line-based file reading endpoint to the workspace agent,
replacing the unbounded byte-based approach for the `read_file` chat
tool and `coder_workspace_read_file` MCP tool.
**Problem**: The current `read_file` tool returns the entire file
contents with no limits, which can blow up LLM context windows and cause
OOM issues with large files.
**Solution**: Inspired by [`coder/mux`](https://github.com/coder/mux)
and [`openai/codex`](https://github.com/openai/codex), implement a
line-based reader with safety limits.
## Changes
### Agent (`agent/agentfiles/`)
- New `/read-file-lines` endpoint with `HandleReadFileLines` handler
- Line-based `offset` (1-based line number, default: 1) and `limit`
(line count, default: 2000)
- Safety constants:
| Constant | Value | Purpose |
|---|---|---|
| `MaxFileSize` | 1 MB | Reject files larger than this at stat |
| `MaxLineBytes` | 1,024 | Per-line truncation with `... [truncated]`
marker |
| `MaxResponseLines` | 2,000 | Max lines per response |
| `MaxResponseBytes` | 32 KB | Max total response size |
| `DefaultLineLimit` | 2,000 | Default when no limit specified |
- Line numbering format: `1\tcontent` (tab-separated)
- Structured JSON response: `{ success, file_size, total_lines,
lines_read, content, error }`
- Hard errors when limits exceeded — tells the LLM to use
`offset`/`limit`
- Existing byte-based `/read-file` endpoint preserved (used by
`instruction.go`)
### SDK (`codersdk/workspacesdk/`)
- `ReadFileLinesResponse` type added
- `ReadFileLines` method added to `AgentConn` interface
- Mock regenerated
### Chat tool (`coderd/chatd/chattool/`)
- `read_file` tool now uses `conn.ReadFileLines()` instead of
`conn.ReadFile()`
- Updated tool description to document line-based parameters
- Response includes `file_size`, `total_lines`, `lines_read` metadata
### MCP tool (`codersdk/toolsdk/`)
- `coder_workspace_read_file` updated to use line-based reading
- Schema descriptions updated for line-based offset/limit
- Removed `maxFileLimit` constant (agent handles limits now)
### Tests
- 13 new test cases for `TestReadFileLines`:
- Path validation (empty, relative, non-existent, directory, no
permissions)
- Empty file handling
- Basic read, offset, limit, offset+limit combinations
- Offset beyond file length
- Long line truncation (>1024 bytes)
- Large file rejection (>1MB)
- All existing tests pass unchanged
## Design decisions
| Decision | Rationale |
|---|---|
| Line-based, not byte-based | Both coder/mux and openai/codex use
line-based — matches how LLMs reason about code |
| Default limit of 2000 | Matches codex; prevents accidental full-file
dumps while being generous |
| 32 KB response cap | Compromise between mux (16 KB) and codex (no cap)
|
| 1024 byte/line truncation with marker | More generous than codex
(500), marker helps LLM know data is missing |
| Hard errors on overflow | Matches mux; forces LLM to paginate rather
than getting partial data |
| Preserve byte-based endpoint | `instruction.go` needs raw byte access
for AGENTS.md |
## Problem
Chat titles revert to the fallback truncated title after briefly showing
the AI-generated title. Even reloading the page doesn't help — the
correct title flashes then gets overwritten.
## Root Cause
Single bug, two symptoms.
In `processChat` (`coderd/chatd/chatd.go`), the `chat` variable is
passed by value. The flow:
1. `processChat(ctx, chat)` receives `chat` with the initial fallback
title (truncated first message).
2. Inside `runChat`, `maybeGenerateChatTitle` generates an AI title,
writes it to the DB via `UpdateChatByID`, and publishes a `title_change`
event. **The DB has the correct title.** The client briefly displays it.
3. `runChat` returns. The **deferred cleanup** in `processChat`
publishes `publishChatPubsubEvent(chat, StatusChange)` — but `chat` here
is the original value copy that still has the **old fallback title**.
4. The frontend receives the `status_change` SSE event and
**unconditionally applies `title` from every event kind** (see
`AgentsPage.tsx` line ~305: `title: updatedChat.title`). This overwrites
the correct AI title with the stale fallback.
**Why reload doesn't help:** If the chat is still processing when the
page reloads, `listChats` loads the correct title from the DB, but then
the deferred `status_change` event arrives moments later and clobbers
it. The title was always in the DB — it was the pubsub event that kept
overwriting it.
## Fix
Re-read the chat from the database in the deferred cleanup before
publishing the final `status_change` event, so it carries the current
(AI-generated) title.
When navigating to a specific agent on the Agents page, the browser tab
title now reflects the agent's chat title (e.g. `Fix login bug - Agents
- Coder`). When the title hasn't loaded yet or when navigating away, it
falls back to `Agents - Coder`.
**Changes:**
- Added a `useEffect` in `AgentDetail` that sets `document.title` via
the existing `pageTitle` utility whenever the chat title changes.
- The cleanup function resets the title back to `Agents - Coder` when
unmounting (navigating away from the agent).
When injecting system instructions into the chat prompt, include:
1. **Operating system** and **working directory** from the
`workspace_agents` table
2. **Home-level instructions** from `~/.coder/AGENTS.md` (existing
behavior)
3. **Project-level instructions** from `<pwd>/AGENTS.md` (new)
The XML tag is renamed from `<coder-home-instructions>` to
`<system-instructions>` since it now carries more than just the home
instruction file.
### Example output (both files present)
```xml
<system-instructions>
Operating System: linux
Working Directory: /home/coder/coder
Source: /home/coder/.coder/AGENTS.md
... home instructions ...
Source: /home/coder/coder/AGENTS.md
... project instructions ...
</system-instructions>
```
### Example output (no AGENTS.md files)
```xml
<system-instructions>
Operating System: linux
Working Directory: /home/coder/coder
</system-instructions>
```
### Changes
- **`coderd/chatd/instruction.go`**:
- Renamed types: `homeInstructionContext` → `agentContext`, added
`instructionFile` struct
- Extracted `readInstructionFileAtPath` shared helper
- Added `readWorkingDirectoryInstructionFile` to read `<pwd>/AGENTS.md`
- Replaced `formatHomeInstruction` with `formatInstructions` that
renders both files under `<system-instructions>`
- **`coderd/chatd/chatd.go`**:
- Renamed `resolveHomeInstruction` → `resolveInstructions`; now reads
both home and pwd instruction files
- `resolveAgentContext` returns `agentContext` (renamed from
`homeInstructionContext`)
- pwd file read is skipped gracefully if directory is empty or file
doesn't exist
- **`coderd/chatd/instruction_test.go`**:
- Added `TestReadWorkingDirectoryInstructionFile` (success, not-found,
empty-directory)
- Replaced `TestFormatHomeInstruction` with `TestFormatInstructions`
covering all combinations
- Added ordering test (`AgentContextBeforeFiles`) to verify OS/pwd
appear before file sources
## Summary
The `chattool` `list_templates` tool previously returned all templates
in a single response with no popularity signal. On deployments with many
templates (e.g. 71 on dogfood), this wastes tokens and makes it hard for
the AI to pick the right template for broad user questions.
## Changes
Single file: `coderd/chatd/chattool/listtemplates.go`
- **`page` parameter** — optional, 1-indexed, 10 results per page
- **Popularity sort** — queries
`GetWorkspaceUniqueOwnerCountByTemplateIDs` to get active developer
counts, then sorts descending (most popular first). The DB query returns
templates alphabetically, so this explicit sort is needed.
- **`active_developers`** — included on each template item so the agent
can see the signal
- **Pagination metadata** — `page`, `total_pages`, `total_count` in the
response so the agent knows there are more results
- **Updated tool description** — tells the agent that results are
ordered by popularity and paginated
## Frontend
No frontend changes needed. The renderer already reads `rec.templates`
and `rec.count` from the response — the new fields (`page`,
`total_pages`, `total_count`) are additive and safely ignored.
When switching between chats on the agents page, the scroll position was
preserved from the previous chat instead of resetting to show the most
recent messages.
## Problem
Clicking a different chat in the sidebar loaded the new chat's messages
but kept the scroll container at whatever position the user had scrolled
to in the previous chat. This meant users often landed in the middle of
a conversation instead of at the bottom where the latest messages are.
## Fix
Added a `useEffect` in `AgentDetail` that resets `scrollTop` to `0`
whenever `agentId` changes. The scroll container uses
`flex-col-reverse`, so `scrollTop = 0` corresponds to the bottom (most
recent messages).
Fixes https://github.com/coder/coder/issues/22375
Updates `stringutil.Truncate` to properly handle multi-byte UTF-8
characters.
Adds tests for multi-byte truncation with word boundary.
Created by Mux using Opus 4.6
Resolves cases where the user is entitled to AI Governance but we don't
show them the page because its not enabled. If for some reason the user
doesn't have AI Bridge enabled anymore but still wants to access the old
logs page they now can.
Furthermore, we link to the docs regardless of if they have AI Bridge
enabled, this is inline with our other settings pages.
Replaces the approach in #22061 (with a cleaner `git history`)
This now ensures that we don't attempt to cause a layout shift when the
sidebars pop-in-out of existence (when scroll locking within `radix`).
This element was receiving the provisioner key daemons and then
immediately filtering them. This lead to the default state being a table
with nothing rendered rather than the `<TableEmpty />` as we would
expect.
<img width="1133" height="608" alt="image"
src="https://github.com/user-attachments/assets/229edb00-b108-4ec3-ac2f-33633c3e5760"
/>
This previously let auditors view the page though they can't update
anything. In a different fashion to #22382 the user will be able to see
all of this as they're logged in to the application anyway, we can
simply tell them `Sorry, no access`.
setup-go has been sporadically failing to download Go, and we were advised
by a member of the Go team that downloading Go from `storage.googleapis.com`
is not guaranteed (which is what setup-go <= v5.6.0 does).
Also remove the use-preinstalled-go optimization for Windows runners.
setup-go v6 sets GOTOOLCHAIN=local, which prevents the pre-installed
Go from auto-downloading the toolchain specified in go.mod. The windows
optimization with v5 relied on GOTOOLCHAIN=auto. setup-go uses the runner
cache, which is a different caching path but should serve the same purpose.
This change adds user-facing feedback when opening apps in a new window
fails due to popup blocking, replacing a silent no-op with a clear
recovery message. It improves reliability and supportability across
app-launch flows by helping users immediately understand and fix the
issue.
This was a poor UX decision to have to reload the entire page when a
template got invalidated. Simply now we refetch the data so that things
come across way smooother.
## Description
- Updates `wsbuilder` to return a `BuildError` with
`http.StatusBadRequest` to signify a "validation error" on missing or
invalid parameters
- Adds a short-circuit in `prebuilds.StoreReconciler` to mark presets
for which creating a build returns a "validation error" as "validation
failed" and skip further attempts to reconcile.
- Adds a test to verify the above
- Introduces a new Prometheus metric
`coderd_prebuilt_workspaces_preset_validation_failed` to track the above
Closes: https://github.com/coder/coder/issues/21237
---------
Co-authored-by: Cian Johnston <cian@coder.com>
State updates from setIsPublishingDialogOpen,
setLastSuccessfulPublishedVersion, and navigation were firing after
waitFor resolved, causing sporadic act() warnings and timeouts in the
publish template version tests (or so says Claude Sonnet 4.6).
Fixescoder/internal#1369
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Related to #22367
It was pointed out to me that we actually did regress this mildly by
removing a dividing line in the changes made in #22367, I've restored
this in a better way by taking advantage of `divide-y` and wrapping this
in a proper `<div />`.
<img width="332" height="385" alt="image"
src="https://github.com/user-attachments/assets/2827a9ae-7b54-4c48-aae9-2f6e965e7f8b"
/>
Switch to asserting only on the onChange spy, which is the actual
component contract being tested. Monaco's textarea value is always empty
regardless of model content, so the toHaveValue assertions were
unreliable anyway.
Fixes the new storybook test introduced in #22202
This was a bad smell that was being addressed by the frontend. This type
was generating out to be a `nil`/`null` instead of an empty `License[]`.
Now this returns as an empty array and we can actively check if we have
no licenses with a length of `0`.
This pull-request takes our icons shown in the sidebar tree and shows
them alongside the names of the files in the `Source Code` page of our
templates.
Also does a quick de-mui of this page.
<img width="637" height="345" alt="image"
src="https://github.com/user-attachments/assets/f3013eb6-9572-4d05-a683-10bb99b4e802"
/>
Adds a brief "Structured Logging" section to the [AI Bridge
Setup](https://coder.com/docs/ai-coder/ai-bridge/setup) page documenting
the `--aibridge-structured-logging` /
`CODER_AIBRIDGE_STRUCTURED_LOGGING` flag.
Covers:
- How to enable structured logging (CLI flag, env var, YAML)
- The five `record_type` values emitted (`interception_start`,
`interception_end`, `token_usage`, `prompt_usage`, `tool_usage`) and
their key fields
- How to filter for these records in a logging pipeline
Created on behalf of @dannykopping
---------
Co-authored-by: blink-so[bot] <211532188+blink-so[bot]@users.noreply.github.com>
Fixes three bugs that caused `coder update` to always re-prompt for
multi-select (`list(string)`) parameters instead of reusing previous
build values:
1. **`isValidTemplateParameterOption` failed for multi-select values**
(`cli/parameterresolver.go`): It compared the entire JSON array string
(e.g. `["vim","emacs"]`) against individual option values, which never
matched. Now parses the JSON array and validates each element
separately.
2. **`RichParameter` ignored previous build value for multi-select**
(`cli/cliui/parameter.go`): The `list(string)` branch always used the
template's default value instead of the `defaultValue` argument (which
carries the previous build's value). Now uses `defaultValue` when
available, falling back to the template default.
3. **Pre-existing crash when `list(string)` has no default value**
(`cli/cliui/parameter.go`): `json.Unmarshal` on an empty string caused
`unexpected end of JSON input`. Now skips unmarshaling when the default
source is empty.
Fixes#19956
The sonner migration (https://github.com/coder/coder/pull/22258) shows
validation errors in both the inline form field and a toast. Scoping the
assertion to the form element avoids flaky matches against the toast.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
The Monaco editor wrapper was only calling `onChange` if the template
file has content, but we want to allow saving an empty file.
Fixes#19721
Claude was used to port tests from jest to vitest, and for the stories.
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Kayla はな <mckayla@hey.com>
