GPG emits an "untrusted key" warning when signing with a key that hasn't
been assigned a trust level, which can cause verification steps to fail
or produce noisy output.
Example:
```sh
gpg: Signature made Tue Mar 24 20:56:59 2026 UTC
gpg: using RSA key 21C96B1CB950718874F64DBD6A5A671B5E40A3B9
gpg: Good signature from "Coder Release Signing Key <security@coder.com>" [unknown]
gpg: WARNING: This key is not certified with a trusted signature!
gpg: There is no indication that the signature belongs to the owner.
Primary key fingerprint: 21C9 6B1C B950 7188 74F6 4DBD 6A5A 671B 5E40 A3B9
```
After importing the release key, derive its fingerprint from the keyring
and mark it as ultimately trusted via `--import-ownertrust`.
The fingerprint is extracted dynamically rather than hard-coded, so this
works for any key supplied via `CODER_GPG_RELEASE_KEY_BASE64`.
## Problem
Template administrators cannot delete templates that have running
prebuilds.
The `deleteTemplate` handler fetches all non-deleted workspaces and
blocks
deletion if any exist, making no distinction between human-owned
workspaces
and prebuild workspaces (owned by the system `PrebuildsSystemUserID`).
This forces admins into a manual multi-step workflow: set
`desired_instances`
to 0 on every preset, wait for the reconciler to drain prebuilds, then
retry
deletion. Prebuilds are an internal system concern that admins should
not need
to manage manually.
## Fix
Replace the blanket `len(workspaces) > 0` guard in `deleteTemplate` with
a
loop that only blocks deletion when a non-prebuild (human-owned)
workspace
exists. Prebuild workspaces — owned by `database.PrebuildsSystemUserID`
— are
now ignored during the check.
Once the template is soft-deleted (`deleted=true`), the existing
prebuilds
reconciler detects `isActive()=false` and cleans up remaining prebuilds
asynchronously. No changes to the reconciler are needed.
The error message and HTTP status for human workspaces remain unchanged.
## Testing
Added two new subtests to `TestDeleteTemplate`:
- **`OnlyPrebuilds`**: deletion succeeds when only prebuild workspaces
exist.
- **`PrebuildsAndHumanWorkspaces`**: deletion is blocked when both
prebuild
and human workspaces exist.
Existing reconciler test ("soft-deleted templates MAY have prebuilds")
already
covers post-deletion prebuild cleanup.
> **PR Stack**
> 1. #23351 ← `#23282`
> 2. #23282 ← `#23275`
> 3. **#23275** ← `#23349` *(you are here)*
> 4. #23349 ← `main`
---
## Summary
Extracts a structured error classification subsystem for agent chat
(`chatd`) so that retry and error payloads carry machine-readable
metadata — error kind, provider name, HTTP status code, and retryability
— instead of raw error strings.
This is the **backend half** of the error-handling work. The frontend
counterpart is in #23282.
## Changes
### New package: `coderd/chatd/chaterror/`
Canonical error classification — extracts error kind, provider, status
code, and user-facing message from raw provider errors. One source of
truth that drives both retry policy and stream payloads.
- **`kind.go`**: Error kind enum (`rate_limit`, `timeout`, `auth`,
`config`, `overloaded`, `unknown`).
- **`signals.go`**: Signal extraction — parses provider name, HTTP
status code, and retryability from error strings and wrapped types.
- **`classify.go`**: Classification logic — maps extracted signals to an
error kind.
- **`message.go`**: User-facing message templates keyed by kind +
signals.
- **`payload.go`**: Projectors that build `ChatStreamError` and
`ChatStreamRetry` payloads from a classified error.
### Modified
- **`codersdk/chats.go`**: Added `Kind`, `Provider`, `Retryable`,
`StatusCode` fields to `ChatStreamError` and `ChatStreamRetry`.
- **`coderd/chatd/chatretry/`**: Thinned to retry-policy only;
classification logic moved to `chaterror`.
- **`coderd/chatd/chatloop/`**: Added per-attempt first-chunk timeout
(60 s) via `guardedStream` wrapper — produces retryable
`startup_timeout` errors instead of hanging forever.
- **`coderd/chatd/chatd.go`**: Publishes normalized retry/error payloads
via `chaterror` projectors.
<!--
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.
-->
Adds the AI Bridge sessions list page.
## Summary
The `build` CI job on `main` is failing with:
```
ERROR: Computed invalid windows version format: 2.32.0-rc.0.1
```
This started when the `v2.32.0-rc.0` tag was created, making `git
describe` produce versions like `2.32.0-rc.0-devel+4f571f8ff`.
## Root cause
`scripts/build_go.sh` converts the version to a Windows-compatible
`X.Y.Z.{0,1}` format by stripping pre-release segments. It uses
`${var%-*}` (shortest suffix match), which only removes the last
`-segment`. For RC versions this leaves `-rc.0` intact:
```
2.32.0-rc.0-devel → strip %-* → 2.32.0-rc.0 → + .1 → 2.32.0-rc.0.1 ✗
```
## Fix
Switch to `${var%%-*}` (longest suffix match) so all pre-release
segments are stripped from the first hyphen onward:
```
2.32.0-rc.0-devel → strip %%-* → 2.32.0 → + .1 → 2.32.0.1 ✓
```
Verified all version patterns produce valid output:
| Input | Output |
|---|---|
| `2.32.0` | `2.32.0.0` |
| `2.32.0-devel` | `2.32.0.1` |
| `2.32.0-rc.0-devel` | `2.32.0.1` |
| `2.32.0-rc.0` | `2.32.0.1` |
Fixes
https://github.com/coder/coder/actions/runs/23511163474/job/68434008241
Nine subtests covering the poll loop, pubsub notification path,
timeout, context cancellation, descendant auth check, and both
error-status branches in handleSubagentDone.
Wire p.clock through awaitSubagentCompletion's timer and ticker
so future tests can use quartz mock clock. Tests use channel-based
coordination and context.WithTimeout instead of time.Sleep.
Coverage: awaitSubagentCompletion 0%->70.3%, handleSubagentDone
0%->100%, checkSubagentCompletion 0%->77.8%,
latestSubagentAssistantMessage 0%->78.9%.
Super unclear why CI hates me and only [fails
lint](https://github.com/coder/coder/actions/runs/23504799702/job/68409588632?pr=23385)
for me (I feel personally attacked), but `dpdm` detected a circular
dependency between the WorkspaceSettingPage and its Sidebar in my other
branch. They both wanted the same context/hook combo, so easy solve to
move the hook/context into a third module to resolve the circular dep.
## Summary
Large pasted text that the UI collapses into an attachment chip was
completely invisible to the LLM. Providers only accept specific MIME
types (images, PDFs) in file content blocks — a `text/plain` `FilePart`
is silently dropped, so the model received nothing for pasted content.
## Fix
Detect paste-originated text files by their
`pasted-text-{timestamp}.txt` filename pattern and convert them to
`fantasy.TextPart` with a bounded 128 KiB inline body and truncation
notice. Binary uploads and real uploaded text files keep their existing
`FilePart` semantics.
The detection uses the existing frontend naming convention
(`pasted-text-YYYY-MM-DD-HH-MM-SS.txt`) combined with a text-like MIME
check for defense-in-depth. A TODO marks this for migration to explicit
origin metadata.
<details>
<summary>Review notes: intentionally skipped findings</summary>
A 10-reviewer deep review was run on this change. The following findings
were raised and intentionally dropped after cross-check. Documenting
them here so future reviewers do not re-flag the same concerns:
**"Unresolved file IDs cause silent data loss" (Edge Case Analyst P1)**
— When a file ID is not in the resolver map, `name` stays empty and
paste detection fails. This is pre-existing behavior for ALL file types
(not introduced by this change). The resolver calls `GetChatFilesByIDs`
which returns whatever rows exist; missing IDs simply fall through to an
empty `FilePart`. The Contract Auditor independently traced this path
and confirmed the fallback is safe. If the file was deleted between
message construction and conversion, the model already saw nothing
before this patch — this change does not make it worse.
**"String builder pre-allocation overhead" (Performance Analyst P1)** —
Misidentified scope. `formatSyntheticPasteText` is only called when
`isSyntheticPaste` returns true (actual synthetic pastes), not for every
file part. The `Grow()` call is correct and efficient.
**"Constant naming violates Uber style" (Style Reviewer P1)** —
Over-severity. `syntheticPasteInlineBudget` is standard Go camelCase for
unexported constants, consistent with the Uber guide and surrounding
code.
**"`IsSyntheticPasteForTest` naming is misleading" (Style Reviewer P2)**
— This is the standard Go `export_test.go` pattern. The `ForTest` suffix
is conventional.
</details>
Observations bypassed the severity test entirely. A reviewer filing
a convention violation as Obs meant it skipped both the upgrade
check and the unnecessary-novelty gate. The combination let issues
pass through as dropped observations when they warranted P3+.
Two changes:
- Severity test now applies to findings AND observations.
- Unnecessary novelty check now covers reviewer-flagged Obs.
When developers switch branches, the database may have migrations
from the other branch that don't exist in the current binary.
This causes coder server to fail at startup, leaving developers
stuck.
The develop script now detects this before starting the server:
1. Connects to postgres (starts temp embedded instance for
built-in postgres, or uses CODER_PG_CONNECTION_URL).
2. Compares DB version against the source's latest migration.
3. If DB is ahead, searches git history for the missing down
SQL files and applies them in a transaction.
4. If git recovery fails (ambiguous versions across branches,
missing files), falls back to resetting the public schema.
Also adds --reset-db and --skip-db-recovery flags.
When edit_files receives multiple files, each file was processed
independently: read, compute edits, write. If file B failed, file A
was already written to disk. The caller got an error but had no way
to know which files were modified.
Split editFile into prepareFileEdit (read + compute, no side
effects) and a write phase. The handler runs all preparations
first and writes only if every file's edits succeed.
A write-phase failure (e.g. disk full) can still leave earlier
files committed. True cross-file atomicity would require
filesystem transactions. The prepare phase catches the common
failure modes: bad paths, search misses, permission errors.
## Summary
Fixes several bugs in the markdown URL transform that replaces
`localhost` URLs with workspace port-forward URLs in the AI agent chat.
## Bugs Fixed
### 1. URLs without an explicit port produce `NaN` in the subdomain
When an LLM outputs a URL like `http://localhost/path` (no port),
`parsed.port` is the empty string `""`. `parseInt("", 10)` returns
`NaN`, producing a broken URL like:
```
http://NaN--agent--workspace--user.proxy.example.com/path
```
Now defaults to port 80 for HTTP and 443 for HTTPS via the new
`resolveLocalhostPort()` helper.
### 2. Protocol always hardcoded to `"http"`
The `urlTransform` in `AgentDetail.tsx` always passed `"http"` as the
protocol argument, silently discarding the original URL's scheme. This
meant `https://localhost:8443/...` would not get the `s` suffix in the
subdomain. Now extracts the protocol from the parsed URL, matching the
existing behavior of `openMaybePortForwardedURL`.
### 3. `urlTransform` not memoized
The closure was re-created on every render. Wrapped in `useCallback`
with the four primitive dependencies (`proxyHost`, `agentName`,
`wsName`, `wsOwner`).
### 4. Duplicated `localHosts` definition
The localhost detection set was defined separately in both
`AgentDetail.tsx` and `portForward.ts`. Consolidated into a single
shared export from `portForward.ts`.
## Changes
- **`site/src/utils/portForward.ts`**: Export shared `localHosts` set
and new `resolveLocalhostPort()` helper. Update
`openMaybePortForwardedURL` to use both.
- **`site/src/pages/AgentsPage/AgentDetail.tsx`**: Import shared
`localHosts` and `resolveLocalhostPort`. Fix protocol extraction.
Memoize `urlTransform`.
- **`site/src/utils/portForward.jest.ts`**: Add tests for
`resolveLocalhostPort` and `localHosts`. Renamed from `.test.ts` to
`.jest.ts` to match project convention.
Wraps the chat timeline in React's <Profiler> to emit
performance.measure() entries and throttled console.warn for
slow renders. Inert in standard builds, only produces output
with a profiling build.
Refs #23354
The React Compiler (babel-plugin-react-compiler@1.0.0) handles
memoization automatically for all components in the AgentsPage
compiled path. Three memo() wrappers were redundant:
- ChatMessageItem in ConversationTimeline.tsx
- LazyFileDiff in DiffViewer.tsx
- ChatTreeNode in AgentsSidebar.tsx
Also migrate three Context.Provider usages to the React 19
shorthand (<Context value={...}>) and simplify the EmbedContext
export to use the context directly instead of re-exporting
.Provider as an alias.
Linear's MCP server (`mcp.linear.app`) returns `token_type="bearer"`
(lowercase) in its OAuth2 token response but rejects requests that use
the lowercase form in the `Authorization` header. RFC 6750 says the
scheme is case-insensitive, but Linear enforces capital-B `Bearer`.
Confirmed by running the actual Linear MCP OAuth flow end-to-end:
- `Authorization: Bearer <token>` → **42 tools, works**
- `Authorization: bearer <token>` → **401 invalid_token**
This is a one-line fix: normalize any case variant of `bearer` to
`Bearer` before building the `Authorization` header, matching the
behavior of the mcp-go library's own OAuth handler.
Depot runners are running out of disk space and blocking builds.
Temporarily switch the build and release jobs from depot runners to
GitHub-hosted runners:
- `ci.yaml` build job: `depot-ubuntu-22.04-8` → `ubuntu-latest`
- `release.yaml` check-perms + release jobs: `depot-ubuntu-22.04-8` →
`ubuntu-latest`
**This is intended to be reverted once depot resolves their disk space
issues.**
> 🤖 This PR was created with the help of Coder Agents, and will be
reviewed by my human. 🧑💻
<img width="684" height="540" alt="image"
src="https://github.com/user-attachments/assets/ccd09873-4640-4a54-b3ca-f740dd50b38d"
/>
## Changes
- Move filter dropdown from top nav bar to inline with the first time
group header (e.g. "Today")
- Remove analytics icon from desktop sidebar nav bar
- Change "View details" to "View usage" in the usage indicator dropdown
- Fix green progress bar visibility in dark mode (`bg-surface-green` →
`bg-content-success`)
- Fix missing space before date in "Resets" text
---
PR generated with Coder Agents
- update some config settings to support "absolute"-style imports by
using a `#/` prefix
- migrate some of the imports in the `WorkspacesPage` to use the new
import style as a proof of concept
because of the change in import sorting behavior this results in, this
diff is already kind of hard to look at–even just from a small migration
for a single page. I think breaking this up into bite size pieces isn't
gonna be worth the work, and leaves more time for merge conflicts to
accrue, more times people would likely have to resolve them.
so I think as far as process for this, I'd like to...
- merge this PR as is, where the config changes are relatively easy to
spot in the haystack, with just enough imports updated to prove that the
config changes are correct
- merge another mega PR after this one which just bites the bullet and
migrates everything else in one fell swoop. it'll probably result in a
ton of merge conflicts for open PRs, but at least it'll only do so once
and then it can be over with.
The @ imports at the bottom of this file are auto-loaded by Claude Code
but silently ignored by other agent runtimes (Coder Agents, Zed, etc.).
Add an explicit fallback so those agents know what to read and when.
After a long requestAnimationFrame pause (e.g. backgrounded tab), the
time delta can be very large, causing the character budget to spike and
bypass smooth rendering entirely. Clamp to 100ms.
Extracted from #23236.
## Problem
MCP servers like Linear (`mcp.linear.app`) require PKCE (RFC 7636) for
their OAuth2 flow. Without it, the token exchange may succeed but the
resulting access token is immediately rejected with a 401
`invalid_token` error when the chat daemon tries to connect to the MCP
server.
This means users can authenticate successfully in the UI (the OAuth
popup completes, `auth_connected` shows `true`), but the model never
receives the MCP tools — they silently fail to load.
### Root cause
The `mcpServerOAuth2Connect` handler was calling
`oauth2Config.AuthCodeURL(state)` without any PKCE parameters
(`code_challenge`, `code_challenge_method`). The callback was calling
`oauth2Config.Exchange(ctx, code)` without a `code_verifier`. Linear's
MCP OAuth endpoint decoded state confirms it expected PKCE with
`codeChallengeMethod: "plain"`.
### Investigation
- The chat (`c2c04fc5-5622-4b71-a5a9-80508e86f78e`) had the Linear MCP
server ID in `mcp_server_ids`
- `auth_connected: true` (token row exists in DB)
- No "expired" or "empty token" warnings in logs
- Server log showed: `skipping MCP server due to connection failure ...
error="initialize: transport error: request failed with status 401:
{"error":"invalid_token","error_description":"Missing or invalid access
token"}"`
- Decoding Linear's OAuth state revealed PKCE was expected
## Changes
- Generate a PKCE `code_verifier` during the OAuth2 connect step using
`oauth2.GenerateVerifier()` and store it in a cookie scoped to the
callback path
- Include `code_challenge` (S256) in the authorization redirect URL via
`oauth2.S256ChallengeOption()`
- Pass the `code_verifier` during the token exchange in the callback via
`oauth2.VerifierOption()`
- Fix a nil-pointer guard on `api.HTTPClient` in the callback
- Add tests verifying PKCE parameters are sent correctly and backwards
compatibility when no verifier cookie is present
feat: add deep-review skill for multi-reviewer code review
Add a skill to .agents/skills/deep-review/ that orchestrates parallel
code reviews from domain-specific reviewers (test auditor, security
reviewer, concurrency reviewer, etc.), cross-checks their findings for
contradictions and convergence, then posts a single structured GitHub
review with inline comments.
Each reviewer reads only its own methodology file (roles/{name}.md) to
preserve independent perspectives. The orchestrator cross-checks across
all findings before posting, tracing combined consequences and
calibrating severity in both directions.
Key capabilities: re-review gate for tracking prior findings across
rounds, consequence-based severity (P0-P4), quoting discipline
separating reviewer evidence from orchestrator judgment, and author
independence (same rigor regardless of who wrote the PR).
- Detect `-testify.m` sub-test filtering in `SetupSuite` and skip the `Accounting` check.
> 🤖 This PR was created with the help of Coder Agents, and was reviewed by my human. 🧑💻
## Summary
Stabilizes the /agents chat viewport so users can read older messages
without being yanked to the bottom when new content arrives.
## Architecture
Replaces the implicit scroll-follow behavior with
**ResizeObserver-driven
scroll anchoring**:
- **`autoScrollRef`** is the single source of truth. User scrolling away
from bottom turns it off; scrolling back near bottom or clicking the
button turns it back on.
- A **content ResizeObserver** on an inner wrapper detects transcript
growth. When auto-scroll is on, it re-pins to bottom via double-RAF
(waiting for React commit + layout to settle). When off, it
compensates `scrollTop` by the height delta to preserve the reading
position. Sign-aware for both Chrome-style negative and Firefox-style
positive `flex-col-reverse` scrollTop conventions.
- Compensation is **skipped during pagination** (older messages prepend
into the overflow direction; the browser preserves scrollTop) and
**during reflow** from width changes.
- A **container ResizeObserver** re-pins to bottom after viewport
resizes
(composer growth, panel changes) when auto-scroll is on.
- **`isRestoringScrollRef`** guards against feedback loops from
programmatic scroll writes. The smooth-scroll guard stays active
until the scroll handler detects arrival at bottom.
## Files changed
- **AgentDetailView.tsx**: Rewrote `ScrollAnchoredContainer` with the
new approach.
- **AgentDetailView.stories.tsx**: Refactored `ScrollToBottomButton`
story
scroll helpers into shared utilities.
## Behavior
- **At bottom + new content**: stays pinned, button hidden.
- **Scrolled up + new content**: reading position preserved, no jump.
- **Viewport resize while pinned**: re-pins to bottom.
- Scroll-to-bottom button and smooth scroll still work.
Adds a `propose_plan` tool that presents a workspace markdown file as a
dedicated plan card in the agent UI.
The workflow is: the agent uses `write_file`/`edit_files` to build a
plan file (e.g. `/home/coder/PLAN.md`), then calls `propose_plan(path)`
to present it. The backend reads the file via `ReadFile` and the
frontend renders it as an expanded markdown preview card.
**Backend** (`coderd/x/chatd/chattool/proposeplan.go`): new tool
registered as root-chat-only. Validates `.md` suffix, requires an
absolute path, reads raw file content from the workspace agent. Includes
1 MiB size cap.
**Frontend** (`site/src/components/ai-elements/tool/`): dedicated
`ProposePlanTool` component with `ToolCollapsible` + `ScrollArea` +
`Response` markdown renderer, expanded by default. Custom icon
(`ClipboardListIcon`) and filename-based label.
**System prompt** (`coderd/x/chatd/prompt.go`): added `<planning>`
section guiding the agent to research → write plan file → iterate → call
`propose_plan`.
OpenAI Responses follow-up turns were replaying full assistant/tool
history even when `store=true`, which breaks after reasoning +
provider-executed `web_search` output.
This change persists the OpenAI response ID on assistant messages, then
in `coderd/x/chatd` switches `store=true` follow-ups to
`previous_response_id` chaining with a system + new-user-only prompt.
`store=false` and missing-ID cases still fall back to manual replay.
It also updates the fake OpenAI server and integration coverage for the
chaining contract, and carries the rebased path move to `coderd/x/chatd`
plus the migration renumber needed after rebasing onto `main`.
## Summary
- use React Query's `dataUpdatedAt` as the remote diff cache
invalidation token instead of a component-local counter
- keep the `@pierre/diffs` cache key stable across remounts without a
custom hashing implementation
- preserve targeted coverage for the cache-key helper used by the
/agents remote diff viewer
## Testing
- `cd site && pnpm exec biome check
src/pages/AgentsPage/components/DiffViewer/RemoteDiffPanel.tsx
src/pages/AgentsPage/components/DiffViewer/diffCacheKey.ts
src/pages/AgentsPage/components/DiffViewer/diffCacheKey.test.ts`
- `cd site && pnpm exec vitest run
src/pages/AgentsPage/components/DiffViewer/diffCacheKey.test.ts
--project=unit`
- `cd site && pnpm exec tsc -p .`
Jake Howell