Implements https://linear.app/codercom/issue/AIGOV-285
Follow the structure established in
https://github.com/coder/coder/pull/25203
## Summary
Adds the `user_ai_budget_overrides` table and CRUD API at
`/api/v2/users/{user}/ai/budget`. An override sets a custom per-user
spend cap that supersedes group-budget resolution, attributing spend to
a specific group.
## Schema
```sql
CREATE TABLE user_ai_budget_overrides (
user_id UUID PRIMARY KEY REFERENCES users(id) ON DELETE CASCADE,
group_id UUID NOT NULL REFERENCES groups(id) ON DELETE CASCADE,
spend_limit_micros BIGINT NOT NULL CHECK (spend_limit_micros >= 0),
created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
);
```
## Membership lifecycle
The membership invariant — a user must be a member of the attributed
group, including when that group is "Everyone" — would naturally be
expressed as a composite FK on `(user_id, group_id) →
group_members_expanded(user_id, group_id)`. PostgreSQL doesn't allow
foreign keys to reference views, so enforcement is split across two
mechanisms:
- **Write-time check.** A CHECK constraint on the table
(`user_ai_budget_overrides_must_be_group_member`) calls a `STABLE`
function `is_group_member(user_id, group_id)` that queries
`group_members_expanded`. The view surfaces both regular group
memberships and the implicit "Everyone" group memberships from
`organization_members`. Any INSERT or UPDATE that violates the predicate
is rejected with a Postgres `check_violation`, which the handler maps to
a 400. `is_group_member` is defined as a general predicate, reusable by
any future table that needs the same check.
- **Cascade on removal.** Two `BEFORE DELETE` triggers handle membership
loss:
- `trigger_delete_user_ai_budget_overrides_on_group_member_delete` on
`group_members` — covers regular group removals (admin action, OIDC
sync).
- `trigger_delete_user_ai_budget_overrides_on_org_member_delete` on
`organization_members` — covers the "Everyone" group, whose membership
lives in `organization_members`.
The single-column FKs on `users(id)` and `groups(id)` remain to cascade
on user or group deletion (those paths don't pass through
`group_members`).
## Authorization
The dbauthz layer gates each operation against the `User` and (for
writes) `Group` resources:
| Operation | User resource | Group resource |
|-----------|----------------|----------------|
| `GET` | `ActionRead` | — |
| `PUT` | `ActionUpdate` | `ActionUpdate` |
| `DELETE` | `ActionUpdate` | `ActionUpdate` |
For `DELETE`, the dbauthz layer fetches the existing override first to
learn the attributed `group_id`, then runs both checks.
### Role matrix
| Role | GET | PUT | DELETE |
|--------------|-----|-----|--------|
| Owner | ✅ | ✅ | ✅ |
| UserAdmin | ✅ | ✅ | ✅ |
| OrgAdmin | ✅ | ❌ | ❌ |
| OrgUserAdmin | ✅ | ❌ | ❌ |
Internal discussion:
https://codercom.slack.com/archives/C096PFVBZKN/p1779392747885359
## Audit logs
Audit logs will be addressed in a follow-up PR.
In order to allow Coder Agents to use AI Gateway in OSS, we need to rehome the `aibridged`\-related code into the AGPL path.
The HTTP API is only registered under enterprise so will still require the AI Governance Add-on to be present in order to use it, whereas Coder Agents uses an in-memory pipe to the same handlers.
`CODER_AI_GATEWAY_ENABLED` / `CODER_AIBRIDGE_ENABLED` is now being defaulted to `true` now that it will be used by Coder Agents.
If you previously had this value disabled explicitly, that value will persist.
Closes
https://linear.app/codercom/issue/AIGOV-284/add-group-budgets-table-and-crud-api
## Summary
Adds the `group_ai_budgets` table and the following endpoints:
- `GET /api/v2/groups/{group}/ai/budget`
- `PUT /api/v2/groups/{group}/ai/budget`
- `DELETE /api/v2/groups/{group}/ai/budget`
Each group may have at most one budget row. If no row exists, no budget
is enforced.
### Feature gate
Added `RequireFeatureMW(FeatureAIBridge)` on the `/ai/budget` sub-route.
## RBAC
Authorization reuses `rbac.ResourceGroup` with the existing
`.InOrganization(...).WithID(...)` scoping model.
The `dbauthz` wrappers load the parent `groups` row and authorize
against it.
No new resource type is introduced. As a result, anyone with
`group:update` permissions (Owner, OrgAdmin, or UserAdmin within the
organization) can manage AI budgets for that group.
## Read access for group members
`database.Group.RBACObject()` grants `policy.ActionRead` to all members
of the group through the group ACL:
```go
func (g Group) RBACObject() rbac.Object {
return rbac.ResourceGroup.WithID(g.ID).
InOrg(g.OrganizationID).
// Group members can read the group.
WithGroupACL(map[string][]policy.Action{
g.ID.String(): {
policy.ActionRead,
},
})
}
```
Because the `GET` endpoint authorizes against the same loaded `Group`
object, any group member can call:
```text
GET /api/v2/groups/{group}/ai/budget
```
`PUT` and `DELETE` remain admin-only. The group ACL grants only
`ActionRead`, so write operations continue to require role-based
`group:update` permissions.
## Alternative considered
A dedicated `rbac.ResourceGroupAiBudget` resource would allow budget
management to be separated from general group administration.
We decided not to add that complexity for now.
Previously, the sessions list sorted by `MIN(started_at)` across
interceptions, so sessions with old start times but recent activity
would sink to the bottom of the list regardless of how recently they
were used.
`ListAIBridgeSessions` now sorts by `COALESCE(MAX(prompt.created_at),
MIN(started_at)) DESC`, exposed as the non-nullable `last_active_at`
field. Sessions with prompts surface by last activity; sessions with no
prompts fall back to their start time.
The original implementation used two separate columns (`last_active_at`
as a nullable prompt timestamp and `sort_at` as the non-nullable cursor
key). This revision collapses them into a single `last_active_at` that
is always set — simplifying the SQL, the Go conversion, the API type,
and the frontend.
🤖 Generated with [Claude Code](https://claude.ai/claude-code)
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
This PR merges code from `coder/aibridge` repository into `coder/coder`.
It was split into 4 PRs for easier review but stacked PRs will need to
be merged into this PR so all checks pass.
* https://github.com/coder/coder/pull/24190 -> raw code copy (this PR,
before merging PRs on top of it, it was just 1 commit:
https://github.com/coder/coder/commit/70d33f33200c7e77df910957595715f81f9bec24)
* https://github.com/coder/coder/pull/24570 -> update imports in
`coder/coder` to use copied code
* https://github.com/coder/coder/pull/24586 -> linter fixes and CI
integration (also added README.md)
* https://github.com/coder/coder/pull/24571 -> added exclude to
scripts/check_emdash.sh check
Original PR message (before PR squash):
Moves coder/aibridge code into coder/coder repository.
Omitted files:
- `go.mod`, `go.sum`, `.gitignore`, `.github/workflows/ci.yml,`
`Makefile`, `LICENSE`, `README.md` (modified README.md is added later)
- `.github`, `example`, `buildinfo,` `scripts` directories
Simple verification script (will list omitted files)
```
tmp=$(mktemp -d)
echo "$tmp"
git clone --depth=1 https://github.com/coder/aibridge "$tmp/aibridge"
git clone --depth=1 --branch pb/aibridge-code-move https://github.com/coder/coder "$tmp/coder"
diff -rq --exclude=.git "$tmp/aibridge" "$tmp/coder/aibridge"
# rm -rf "$tmp"
```
## Summary
Adds `--ai-gateway-allow-byok` deployment option to control whether
users can use Bring Your Own Key (BYOK) mode with AI Gateway.
When disabled (`--ai-gateway-allow-byok=false`), BYOK requests are
rejected with a 403 and a message directing the admin to enable the
flag. Centralized key authentication works regardless of this setting.
Defaults to `true` (BYOK allowed).
---------
Co-authored-by: Danny Kopping <danny@coder.com>
## Summary
Adds a `sanitizeCredentialHint` safety check in the db-to-SDK conversion
layer to ensure credential hints are always masked before being exposed
in the API. Also adds `credential_kind` and `credential_hint` assertions
to the session threads API test.
Two new columns added to aibridge_token_usages:
- cache_read_input_tokens (BIGINT, default 0)
- cache_write_input_tokens (BIGINT, default 0)
Migration backfills existing rows by extracting values from the metadata
JSONB column (cache_read_input, input_cached, prompt_cached for reads
(max value selected since only 1 should be set), cache_creation_input
for writes).
All references to data from metadata were updated to reference new
columns. No other changes then changing where data is extracted from.
Requires aibridge library version bump to include:
https://github.com/coder/aibridge/pull/229
Fixes: https://github.com/coder/aibridge/issues/150
_Disclaimer: produced using Claude Opus 4.6, reviewed by me, and
validated against Dogfood dataset._
The `ListAIBridgeSessions` query materialized and aggregated all
matching interceptions before paginating, then ran expensive
token/prompt lookups across the full dataset. For a page of 25 sessions
against ~200k interceptions (our dogfood dataset), this meant:
- Three CTEs scanning all rows (filtered_interceptions, session_tokens,
session_root)
- ARRAY_AGG(fi.id) collecting every interception ID per session
- Lateral prompt lookup via ANY(array_of_all_ids) running for every
session, not just the page
- ~90MB of disk sorts and JIT compilation kicking in
The improvement is to restructure to paginate first and enrich after: a
single CTE groups interceptions into sessions with only cheap aggregates
(MIN, MAX, COUNT), applies cursor pagination and LIMIT, then lateral
joins fetch metadata, tokens, and prompts for just the ~25-row page.
Measured against 220k interceptions / 160k sessions:
| Metric | Before | After |
|--------------------|--------|-------|
| Execution time | 1800ms | 185ms |
| Shared buffer hits | 737k | 2.6k |
| Disk sort spill | 86MB | 16MB |
| Lateral loops | 160k | 25 |
https://grafana.dev.coder.com/goto/fbODPGtvR?orgId=1 the results are
identical, just _much_ faster.
---
Also includes some additional tests which I added prior to refactoring
the query to ensure no regressions on edge-cases.
---------
Signed-off-by: Danny Kopping <danny@coder.com>
Closes#22136
This pull-request implements a `<ClientFilter />` to our `Request Logs`
page for AI Bridge. This will allow the user to select a client which
they wish to filter against. Technically the backend is able to actually
filter against multiple clients at once however the frontend doesn't
currently have a nice way of supporting this (future improvement).
<img width="1447" height="831" alt="image"
src="https://github.com/user-attachments/assets/0be234e2-25f2-4a89-b971-d74817395da1"
/>
---------
Co-authored-by: Jeremy Ruppel <jeremy.ruppel@gmail.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
_Disclaimer:_ _produced_ _by_ _Claude_ _Opus_ _4\.6,_ _reviewed_ _by_ _me._
**This is a breaking change.** Users who are not have `owner` or sitewide `auditor` roles will no longer be able to view interceptions.
Regular users should not need to view this information; in fact, it could be used by a malicious insider to see what information we track and don't track to exfiltrate data or perform actions unobserved.
---
Changed authorization for AI Bridge interception-related operations from system-level permissions to resource-specific permissions. The following functions now authorize against `rbac.ResourceAibridgeInterception` instead of `rbac.ResourceSystem`:
- `ListAIBridgeTokenUsagesByInterceptionIDs`
- `ListAIBridgeToolUsagesByInterceptionIDs`
- `ListAIBridgeUserPromptsByInterceptionIDs`
Updated RBAC roles to grant AI Bridge interception permissions:
- **User/Member roles**: Can create and update AI Bridge interceptions but cannot read them back
- **Service accounts**: Same create/update permissions without read access
- **Owners/Auditors**: Retain full read access to all interceptions
Removed system-level authorization bypass in `populatedAndConvertAIBridgeInterceptions` function, allowing proper resource-level authorization checks.
Updated tests to reflect the new permission model where members cannot view AI Bridge interceptions, even their own, while owners and auditors maintain full visibility.
<!--
If you have used AI to produce some or all of this PR, please ensure you have read our [AI Contribution guidelines](https://coder.com/docs/about/contributing/AI_CONTRIBUTING) before submitting.
-->
_Disclaimer:_ _initially_ _produced_ _by_ _Claude_ _Opus_ _4\.6,_ _heavily_ _modified_ _and_ _reviewed_ _by_ _me._
Closes https://github.com/coder/internal/issues/1360
Adds a new `/api/v2/aibridge/sessions` API which returns "sessions".
Sessions, as defined in the [RFC](https://www.notion.so/coderhq/AI-Bridge-Sessions-Threads-2ccd579be59280f28021d3baf7472fbe?source=copy_link), are a set of interceptions logically grouped by a session key issued by the client.
The API design for this endpoint was done in [this doc](https://github.com/coder/internal/issues/1360).
If the client has not provided a session ID, we will revert to the thread root ID, and if that's not present we use the interception's own ID (i.e. a session of a single interception - which is effectively what we show currently in our `/api/v2/aibridge/interceptions` API).
The SQL query looks gnarly but it's relatively simple, and seems to perform well (~200ms) even when I import dogfood's `aibridge_*` tables into my workspace. If we need to improve performance on this later we can investigate materialized views, perhaps, but for now I don't think it's warranted.
---
_The PR looks large but it's got a lot of generated code; the actual changes aren't huge._
## Summary
AI Bridge is moving to General Availability in v2.30 and will require
the AI Governance Add-On license in future versions. This adds a soft
warning for deployments using AI Bridge via Premium/Enterprise
FeatureSet without an explicit AI Bridge add-on license.
Relates to: https://github.com/coder/internal/issues/1226
## Changes
- Track whether AI Bridge was explicitly granted via license Features
(add-on) vs inherited from FeatureSet
- Show soft warning when AI Bridge is enabled and entitled via
FeatureSet but not via explicit add-on
- Changed AI Bridge enablement from hardcoded `true` to check
`CODER_AIBRIDGE_ENABLED` deployment config
## Behavior Change
AI Bridge is now only marked as "enabled" in entitlements when
`CODER_AIBRIDGE_ENABLED=true` is set in the deployment config.
Previously, it was always enabled for Premium/Enterprise licenses
regardless of the config setting.
This change ensures that users who do not use AI Bridge will not see the
soft warning about the upcoming license requirement.
## Warning Message
> AI Bridge is now Generally Available in v2.30. In a future Coder
version, your deployment will require the AI Governance Add-On to
continue using this feature. Please reach out to your account team or
sales@coder.com to learn more.
## Behavior
| Condition | Warning Shown |
|-----------|---------------|
| AI Bridge disabled | ❌ No |
| AI Bridge enabled + explicit add-on license | ❌ No |
| AI Bridge enabled + Premium/Enterprise FeatureSet (no add-on) | ✅ Yes
|
## Screenshots
### 1. No license
<img width="1708" height="577" alt="image"
src="https://github.com/user-attachments/assets/cbdbfd4d-55de-4d70-8abf-2665f458e96f"
/>
### 2. No license + CODER_AIBRIDGE_ENABLED=true
<img width="1716" height="513" alt="image"
src="https://github.com/user-attachments/assets/344aae76-7703-485f-b568-1f13a1efa48f"
/>
### 3. Premium license + CODER_AIBRIDGE_ENABLED=false
<img width="1687" height="389" alt="image"
src="https://github.com/user-attachments/assets/c2be12b0-1c0f-438d-a293-f9ec9fe6a736"
/>
### 4. Premium license + CODER_AIBRIDGE_ENABLED=true
<img width="1707" height="525" alt="image"
src="https://github.com/user-attachments/assets/1a4640e1-e656-4f9b-bed0-9390cb5d6a84"
/>
## Notes
- TODO comments added to mark code that should be removed when AI Bridge
enforcement is added
- Feature continues to work - this is just a transitional warning (soft
enforcement)
## Summary
This adds configurable overload protection to the AI Bridge daemon to
prevent the server from being overwhelmed during periods of high load.
Partially addresses coder/internal#1153 (rate limits and concurrency
control; circuit breakers are deferred to a follow-up).
## New Configuration Options
| Option | Environment Variable | Description | Default |
|--------|---------------------|-------------|---------|
| `--aibridge-max-concurrency` | `CODER_AIBRIDGE_MAX_CONCURRENCY` |
Maximum number of concurrent AI Bridge requests. Set to 0 to disable
(unlimited). | `0` |
| `--aibridge-rate-limit` | `CODER_AIBRIDGE_RATE_LIMIT` | Maximum number
of AI Bridge requests per second. Set to 0 to disable rate limiting. |
`0` |
## Behavior
When limits are exceeded:
- **Concurrency limit**: Returns HTTP `503 Service Unavailable` with
message "AI Bridge is currently at capacity. Please try again later."
- **Rate limit**: Returns HTTP `429 Too Many Requests` with
`Retry-After` header.
Both protections are optional and disabled by default (0 values).
## Implementation
The overload protection is implemented as reusable middleware in
`coderd/httpmw/ratelimit.go`:
1. **`RateLimitByAuthToken`**: Per-user rate limiting that uses
`APITokenFromRequest` to extract the authentication token, with fallback
to `X-Api-Key` header for AI provider compatibility (e.g., Anthropic).
Falls back to IP-based rate limiting if no token is present. Includes
`Retry-After` header for backpressure signaling.
2. **`ConcurrencyLimit`**: Uses an atomic counter to track in-flight
requests and reject when at capacity.
The middleware is applied in `enterprise/coderd/aibridge.go` via
`r.Group` in the following order:
1. Concurrency check (faster rejection for load shedding)
2. Rate limit check
**Note**: Rate limiting currently applies to all AI Bridge requests,
including pass-through requests. Ideally only actual interceptions
should count, but this would require changes in the aibridge library.
## Testing
Added comprehensive tests for:
- Rate limiting by auth token (Bearer token, X-Api-Key, no token
fallback to IP)
- Different tokens not rate limited against each other
- Disabled when limit is zero
- Retry-After header is set on 429 responses
- Concurrency limiting (allows within limit, rejects over limit,
disabled when zero)
Addresses [`aibridge#54`](https://github.com/coder/aibridge/issues/54)
When querying against the values in the database for
`/api/experimental/aibridge/interceptions` we found strange behaviour
wherein there was interceptions that lacked prompting and other various
fields we want. Generally this was as a result of the data not actually
existing for these values (as they were inflight).
The simple solution to this was to hide them if they didn't exist. This
PR addresses that.
---------
Co-authored-by: Danny Kopping <danny@coder.com>
Thanks to the great work in #20393, we’ve successfully introduced
offset-based pagination for this endpoint. However, the frontend expects
a `count` field in the response rather than `total`. This PR updates the
response payload to rename the returned key to `count` for consistency
with frontend expectations and existing API patterns.
This is necessary to unblock the work in #20331
- Adds FK from `aibridge_interceptions.initiator_id` to `users.id`
- This is enforced by deleting any rows that don't have any users. Since
this is an experimental feature AND coder never deletes user rows I
think this is acceptable.
- Adds `name` as a property on `codersdk.MinimalUser`
- This matches the `visible_users` view in the database. I'm unsure why
`name` wasn't already included given that `username` is.
- Adds a new `initiator` field to `codersdk.AIBridgeInterception` which
contains `codersdk.MinimalUser` (ID, username, name, avatar URL)
- Removes `initiator_id` from `codersdk.AIBridgeInterception`
- Should be fine since we're still in early access
Necessary for the frontend to be able to paginate easily. Cursor
pagination is good for fetching all events, but doesn't play very well
when a pagination component gets involved.
Adds support for `?offset=x` to the existing endpoint. The cursor-based
pagination (`?after_id=x`) is still supported. The two pagination modes
are mutually exclusive, and are documented as such. If both are
supplied, the request will be rejected.
Also adds a `total` property to the response that contains the full
count of items matching the filter. We already have indices in place so
I don't think this will impact performance (or we can revisit it before
GA).
Yevhenii Shcherbina