Previously, user-uploaded PDFs were silently dropped by fantasy's
Anthropic provider adapter, so Claude (direct or via Bedrock) only saw
the user's text and replied as if no document had been attached. Other
providers (OpenAI, Gemini, OpenRouter, Vercel) were unaffected.
Bumps `coder/fantasy` past
[coder/fantasy#37](https://github.com/coder/fantasy/pull/37)
(cherry-pick of upstream
[charmbracelet/fantasy#197](https://github.com/charmbracelet/fantasy/pull/197)),
which emits an Anthropic `document` content block with a base64 PDF
source for `fantasy.FilePart{MediaType: "application/pdf"}` and counts
`OfDocument` as user-visible so a PDF-only user message is no longer
culled as empty.
Adds a regression test
(`TestModelFromConfig_AnthropicPDFFilePartReachesProvider`) that drives
a `fantasy.FilePart` through the real Anthropic provider against a
`chattest.NewAnthropic` stub and asserts the outbound request contains a
base64 document block. The test was verified to fail on the previous
fantasy pin (the request leaves with zero messages and `Generate`
returns EOF) and pass on the new one.
Manually verified end-to-end with `./scripts/develop.sh`: uploading a
PDF to a Claude-backed Coder Agents chat now lets the model read it.
Closes CODAGT-540
`TestAdvisorHappyPath_RootChat` could subscribe after the active test
server had already processed the chat and published transient advisor
deltas, leaving the live delta collector empty.
Use a passive chatd test server until the live subscriber and collector
are registered, then start processing and wait for the expected advisor
deltas before canceling the stream.
Closescoder/internal#1548
Generated by Coder Agents.
<details>
<summary>Implementation notes</summary>
The failing assertion covered stream-only advisor `ResultDelta` events.
`CreateChat` signals the processor, so an already-started server can
publish those deltas before `Subscribe` registers its local stream
subscriber. The test now creates the chat on a passive server,
subscribes, starts the collector, then calls `Start()`.
</details>
> Mux updated this PR on behalf of Mike.
AI Gateway chat retries after context compaction could lose active turn
API key routing metadata because the prompt query keeps the compressed
model-only summary but omits the original visible user turn.
Persist the active API key ID onto compaction summaries explicitly.
Model construction now uses one active-turn lookup helper for visible
user turns and compressed summary boundaries, so prompt model
construction can recover the key when no later visible user turn exists.
Added unit and DB-backed coverage for the compacted prompt path.
Retry Coder Agents workspace creation once with a generated random
suffix when the requested workspace name already exists. This preserves
structured errors for other conflicts and avoids surfacing avoidable
name collisions.
Closes CODAGT-386
Use testing.Testing() inside createTransport to automatically
clone http.DefaultTransport when running in tests. In production,
DefaultTransport is used as-is (efficient connection pooling).
This fixes the CloseIdleConnections flake class: httptest.Server.Close()
calls http.DefaultTransport.CloseIdleConnections(), which disrupts
any MCP client sharing that transport. The testing.Testing() check
means every MCP transport created during tests gets isolation
automatically, with no caller changes needed.
Closescoder/internal#1016
Closes PLAT-291
Two additions to the compaction summary prompt:
1. Error specificity: the "errors encountered" bullet now instructs the
model to keep error notes specific (name the file, the error, the
fix) and not generalize from a specific failure to a blanket
tool-avoidance rule. This addresses the doctrine crystallization
pattern where a single tool failure gets promoted to a standing
"avoid tool X" rule that persists across compactions and model swaps.
2. Reproducibility: a new closing sentence instructs the model to
reference reproducible content by path, command, or URL rather than
inlining it. Content without a stable reproducer is still preserved
inline with a brief summary. This targets summary bloat from
inlined code blocks (worst case: 34k chars, 76 code blocks
reproducing repo content verbatim).
Refs CODAGT-331
Builds on top of https://github.com/coder/coder/pull/25794
Adds a new `provider_disabled` error classification in `chatd` with the
corresponding plumbing to classify it as non-retryable. Also adds a
story for how this particular error kind is displayed in the UI.
Replaces the 60 second first-token timeout in the chat loop with a 10
minute stream-silence timeout.
Previously, the guard bounded only the gap before the first stream part.
Once any part arrived the attempt could hang indefinitely if the
provider stopped streaming without closing the connection, and even
normal long-running responses could be killed after 60 seconds if the
provider was slow to emit the first token.
The guard now arms when a model attempt opens its stream, resets on
every received stream part, and fires after 10 minutes of complete
silence. The existing retry path still handles the timeout, and the
public `startup_timeout` error kind is preserved to avoid API and
frontend churn.
10 minutes matches the default request timeout used by the Anthropic and
OpenAI Python SDKs.
Closes CODAGT-493
OpenAI-compatible chat paths hit two provider compatibility issues. Some
compatible endpoints reject a named `tool_choice` when there is only one
tool, and Gemini's OpenAI-compatible endpoint requires thought
signatures on current-turn tool calls.
Centralize OpenAI-compatible request patches in the chat provider:
rewrite single named tool choices to `"required"`, and add the
documented dummy Google thought signature to the first tool call in each
current-turn tool step for Gemini routes. Vercel OpenAI-compatible
requests are left unchanged for the thought-signature patch.
> Mux created this PR on behalf of Mike.
- Empty string is valid for `apiKeyID` in paths that genuinely lack a
caller key (e.g. agent-initiated context injection in
`workspaceAgentAddChatContext`). AI Gateway fail-closed check remains
the runtime safety net.
- Context injection paths (`persistInstructionFiles`, compaction) read
the key from `aibridge.DelegatedAPIKeyIDFromContext(ctx)`, set upstream
by `contextWithActiveTurnAPIKeyID`.
- Subagent context copy branches on `copiedRole ==
database.ChatMessageRoleUser` to choose the right append function.
> Generated by Coder Agents
Fixes CODAGT-503
- Add failing-first coverage for manual title generation with missing
message `api_key_id`, with both context fallback and fail-closed cases.
- Set `aibridge.WithDelegatedAPIKeyID(ctx, apiKey.ID)` in
`regenerateChatTitle` and `proposeChatTitle`.
- In `generateManualTitleCandidate`, fall back to
`aibridge.DelegatedAPIKeyIDFromContext(ctx)` only when
`modelBuildOptionsFromMessages` yields an empty `ActiveAPIKeyID`.
- Keep `modelBuildOptionsFromMessages` pure and leave automatic title
generation unchanged.
For `vercel`, `openrouter`, and `openai-compat`, the
`<provider>/<model>` slash is part of the upstream model ID rather than
a hint. `ResolveModelWithProviderHint` was running
`parseCanonicalModelRef` before honoring `providerHint`, so a config
like `(provider=vercel, model=anthropic/claude-4-5-sonnet)` resolved to
`provider=anthropic, model=claude-4-5-sonnet` and the prefix-less model
name was forwarded to Vercel, which returned `Model 'claude-4-5-sonnet'
not found`.
Honor an explicit gateway provider hint before attempting canonical-ref
parsing. Non-gateway hints (anthropic, openai, etc.) keep the existing
canonical-ref-first behavior so `anthropic/claude-...` still has its
prefix stripped when routed directly to Anthropic.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes CODAGT-484.
- Removed "quota", "billing", "insufficient_quota", "payment required"
from `authStrongPatterns`
- Added `usageLimitPatterns` slice with those patterns
- Added `usageLimitMatch` signal and rule between overloaded and
authStrong in priority
- Added terminal/retry messages for `ChatErrorKindUsageLimit`
- Simplified auth message (removed billing reference)
- Frontend: conditional `!usageLimitStatus.provider` guard on the "View
Usage" Alert
- Added `TestClassify_UsageLimitBeatsAuth` with 5 cases including real
production OpenAI error
- Added `ProviderQuotaExceeded` story asserting no "View Usage" link and
correct `ChatStatusCallout` rendering
> Generated with [Coder Agents](https://coder.com/agents)
## Summary
Routes chatd model calls backed by concrete AI Provider rows through the
in-process aibridge transport by default, with deployment options to use
direct provider routing when AI Gateway is disabled or chat AI Gateway
routing is disabled.
- Splits model routing into common, direct provider, and AI Gateway
paths behind a single deployment-mode entry point.
- Builds chatd models through explicit request, route, and options data.
Active API key attribution is passed explicitly instead of being hidden
inside generic model construction.
- For AI Gateway BYOK routes, resolves the user's provider key in chatd,
forwards it through provider-specific auth headers, and sets
`X-Coder-AI-Governance-Token` to the `delegated` marker so aibridge
preserves those headers while still stripping Coder-specific metadata.
- Keeps central provider credentials and deployment fallback credentials
out of forwarded provider auth headers, so AI Gateway central policy
remains authoritative.
- Redacts delegated provider auth from default string formatting to
avoid accidental plaintext logging of user BYOK credentials.
- Covers selected chat models, advisor overrides, title and quickgen
paths, subagent overrides, computer use model selection, and an
integration-style chat turn through the aibridge transport path.
- Persists initiating API key IDs on chat and queued user messages,
including subagent child messages, and fails closed for AI
Gateway-routed model builds without an active key.
- Removes unused `api_key_id` indexes while keeping the persistence
columns and foreign keys.
- Keeps the deployment option available through config and env parsing,
but hides it from CLI help and generated docs.
- Stabilizes the subagent poll fallback test so background CreateChat
processing cannot win the state transition under slower CI environments.
## Tests
- `go test ./coderd/x/chatd -run
'TestAIGatewayProviderAuthForUser|TestAIGatewayProviderAuthRedactsFormatting|TestResolveModelRouteForConfigAIGatewayProviderAuth|TestAIGatewayModelForwardsProviderAuth|TestProcessChat_AIGatewayRoutingUsesDelegatedAPIKey|TestAwaitSubagentCompletion'
-count=1`
- `go test ./coderd/aibridged -run
'TestServeHTTP_DelegatedAPIKey|TestServeHTTP_StripCoderToken' -count=1`
- `git diff --check HEAD~1..HEAD`
- `make lint`
> Mux working on behalf of Mike.
Models frequently confuse the search and replace fields in the
edit_files tool (CODAGT-312). Rename the model-facing JSON fields
to old_text/new_text so the intent is unambiguous.
Backend: custom UnmarshalJSON on editFileEdit falls back to
deprecated search/replace when old_text/new_text are empty. The
workspace agent API is unchanged; toSDKFiles maps old_text/new_text
back to search/replace for agent/agentfiles.
Frontend: normalizeEdit in parseEditFilesArgs accepts both
old_text/new_text and search/replace, normalizing to the internal
{ search, replace } representation so streaming diff rendering
works with either field naming convention.
This fixes the flaky `TestSubscribeAfterMessageID` by seeding its chat
and messages directly, so the test no longer creates pending work that a
chat worker can pick up. The assertion now covers only the
`afterMessageID` subscription behavior, independent of chat processing
lifecycle timing.
Closes DEVEX-326
Closes https://github.com/coder/internal/issues/1489
TestPromoteQueuedWhileRunningRespectsMessageOrder was flaky because
it read queue state from the database immediately after PromoteQueued
returned. The active server worker drains queued messages concurrently,
so the DB read races the auto-promote pipeline (TOCTOU).
Instead of asserting intermediate queue state, wait for all three
promoted messages to appear in chat history and verify their relative
order (B before A before C). This asserts the same invariant (promote
reorders B to the front) without reading during the race window.
Closes CODAGT-384
The root cause of the TestPromoteQueuedWhileRequiresActionMixedTools
flake (CODAGT-425) was the subscriber out-of-order durable message
delivery bug, fixed by PR #25433 (ec1e861). All five CI failures
predate that fix. Zero failures since.
This change hardens the subscriber event-drain pattern in both
PromoteQueued requires_action tests: wrap the channel select in a
for-loop so interleaved non-target events (status, queue_update,
message_parts) are consumed in the same Eventually tick instead of
each burning a 25ms interval. This is defense-in-depth for slow CI
runners, not a standalone bug fix.
Closescoder/internal#1523
Closes CODAGT-425
Skips `TestSignalWakeSendMessage`, which flakes because the current
chatd control notification flow can deliver stale status notifications
after a new processing run starts.
This mirrors the existing CODAGT-353 skips for the same
stale-notification class and leaves the deterministic fix to that
notification-flow refactor.
Refs
https://linear.app/codercom/issue/ENG-2727/flake-testsignalwakesendmessage
> Generated by Coder Agents on behalf of @ibetitsmike.
GPT-class chat turns could eagerly create workspaces or repeat setup
such as cloning an existing repo because the system prompt framed setup
work as the default path.
This updates chatd prompt guidance and the `create_workspace` tool
description so agents reuse existing chat and workspace context, treat
injected workspace context as already read, avoid recloning present
repositories, and create or start workspaces only when workspace-backed
work is required. Delegated chats now report workspace needs to the
parent instead of trying to create one.
> Mux opened this PR on behalf of Mike.
My agent added `//nolint:testpackage` to a test file on one of my PRs.
Again. This PR cleans it up across the entire repo and updates the
in-repo conventions so future agents stop doing it.
The repo already has a precedent for white-box tests that need to touch
unexported symbols: `*_internal_test.go` (145+ existing files). The
`testpackage` linter's default `skip-regexp` exempts that filename
suffix, so the `//nolint:testpackage` directive is unnecessary in every
case where someone reached for it. This PR renames 51 such files to
`*_internal_test.go` via `git mv` so blame and history follow, and
strips the dead directive from 2 files that were already correctly named
(`coderd/oauth2provider/authorize_internal_test.go`,
`coderd/x/chatd/advisor_internal_test.go`).
`.claude/docs/TESTING.md` now documents the rule explicitly under *Test
Package Naming*, which is imported into the root `AGENTS.md` via
`@.claude/docs/TESTING.md`. The rule: prefer `package foo_test`; if you
need internal access, rename the file to `*_internal_test.go` rather
than adding a nolint directive.
> Mux updated this PR on behalf of Mike.
## Summary
- Add experimental personal skills API helpers and an Agents settings UI
for listing, creating, editing, deleting, and importing SKILL.md
content.
- Add docs, Storybook coverage, and unit tests for backend-compatible
SKILL.md parsing.
- Address review feedback by simplifying frontmatter scalar parsing,
clarifying the UI parser scope, defaulting personal skill queries to
`me`, and patching React Query caches after create, update, and delete.
- Merge latest `main` and resolve the Agents sidebar refactor conflicts.
## Validation
- pre-commit hook
- `go test ./codersdk/workspacesdk -run TestParseSkillFrontmatter
-count=1`
- `go test ./coderd/x/chatd/chattool -run 'Test' -count=1`
- `cd site && pnpm test --
src/pages/AgentsPage/utils/personalSkills.test.ts
src/api/queries/userSkills.test.ts src/utils/fileSize.test.ts
--runInBand`
- `cd site && pnpm lint:types`
- `cd site && pnpm lint:check`
> Mux is opening this PR on behalf of Mike.
Agents could interpret a generic "commit and push" request on `main` as
permission to commit on the current branch and push its upstream.
Add version-control safety guidance to the default agent system prompt
so agents check the current branch and push target, avoid default or
protected branch commits and pushes unless the user explicitly confirms
after a warning, avoid plain git push from those branches, and create a
feature branch first when no explicit confirmation is present.
When the execute tool runs a chained shell command, the UI previously
rendered the raw string. Long chains like "cd /repo && git pull &&
git add . && git commit -m fix" were hard to scan.
A new ChatMessagePart.ParsedCommands [][]string field on tool-call
parts carries one entry per simple command, parsed in chatd from args
via mvdan.cc/sh/v3/syntax. The frontend renders the joined list ("cd,
git pull, git add, git commit") in place of the raw command, and falls
back to the raw command when the field is absent.
Closes CODAGT-446
The subscriber advanced a single delivery cursor on each notify and
trusted it for both lookups. Concurrent publishMessage calls and PG
NOTIFY commit ordering let cache appends and notifies arrive out of
ID order, after which a late notify would scan above its own message
and drop it. The DB fallback was also skipped whenever the cache
delivered anything, hiding cross-replica messages that only the DB
held.
The cursor becomes a high-water mark, not the lookup key. Notifies
trigger a rescan over the gap they describe and dedupe per
subscription, and the DB pass runs every time so cross-replica
messages can't get eaten by a local cache hit.
Closescoder/internal#1525
Closes CODAGT-357
> Mux updated this PR on behalf of Mike.
## Stack Context
This PR builds on #25365 in the experimental personal skills stack and
completes the chat integration.
Stack order:
1. #25362 personal skill resolver
2. #25363 storage, permissions, API, and SDK
3. #25365 API test coverage
4. #25366 chattool and chatd integration
5. #25066 settings UI and docs
6. #25386 personal skills slash menu
## What?
Updates chattool skill formatting and `read_skill` resolution so tools
can read personal skills from the database, then injects personal skill
metadata into chatd prompts and registers the skill-reading tools when
skills are available.
This branch has also been merged with current `origin/main` to resolve
merge conflicts.
## Why?
The chattool and chatd changes need to land together so the intermediate
stack state stays buildable. This completes personal skill availability
in chats without syncing personal skills into workspace filesystems.
## Validation
- `go test -count=1 ./coderd/x/chatd/chattool -run
'TestFormatResolvedSkillIndex|TestReadSkillTool|TestReadSkillFileTool'`
- `go test -count=1 ./coderd/x/chatd -run
'TestPersonalSkillsInSystemPrompt|TestPersonalAndWorkspaceSkillCollisionInSystemPrompt|TestSkillIndexRefreshReplacesStaleAliases|TestFetchPersonalSkillMetadata|TestLoadPersonalSkillBody'`
- `go test -count=1 ./coderd -run 'Test.*UserSkill'`
- `git diff --cached --check`
- `make lint`
- pre-commit hook
> Mux prepared this PR on behalf of Mike.
## Stack Context
This is PR 1 of 6 in the `mike/ai-providers` Graphite stack. The stack
migrates Agents chat provider configuration from legacy chat provider
tables to the unified AI provider tables used by the AI provider
administration surface.
See the stack comment for review order and links.
## What?
Adds a package-level `coderd/x/chatd/docs.go` glossary for AI Providers,
provider-scoped keys, user BYOK keys, and Agents as the consuming
feature area.
## Why?
Keeping the glossary next to chatd makes the migration language visible
where Agents consume AI Providers, without adding a separate PRD, root
context file, or ADR structure.
> Mux updated this PR on behalf of Mike.
## Stack Context
This PR is the storage, permissions, API, and SDK layer for experimental
personal skills. #25362 has landed on `main`, so this branch is
restacked directly on `main`.
Stack order:
1. #25363 storage, permissions, API, and SDK
2. #25365 API test coverage
3. #25366 chattool and chatd integration
4. #25066 settings UI and docs
5. #25386 personal skills slash menu
## What?
Adds the `user_skills` database table, generated queries, RBAC resources
and scopes, audit resource handling, experimental user-scoped CRUD
endpoints, SDK types, and generated API/site types.
Follow-up review and restack fixes:
- Enforce a bounded personal skill description in parser and database
constraints.
- Return `403 Forbidden` for unauthorized create and update attempts.
- Return explicit conflict responses when soft-deleted users are
targeted.
- Keep user admins out of personal skills, while site owners can read
and delete but not create or update.
- Document trigger-raised constraint names and keep schema constants
covered by tests.
- Reuse `UserSkillMetadata` in the full `UserSkill` SDK response type.
- Generate user skill IDs in Go instead of relying on a database
default.
- Rebase on latest `main` and renumber the user skills migration to
`000502_user_skills`.
## Why?
Personal skills need durable user-owned storage with owner
authorization, limited site-owner moderation, and a hidden API surface
before chatd can consume them.
## Validation
- `make gen`
- `go test ./coderd/database -run '^TestUserSkillSchemaConstants$'
-count=1`
- `go test ./coderd/database/dbauthz -run
'^TestMethodTestSuite/TestUserSkills$' -count=1`
- `go test ./coderd -run '^TestPatchUserSkill$' -count=1`
- `go test ./codersdk ./coderd/database/db2sdk`
- `make lint`
- pre-commit hook on `97fd58108d`
> Mux opened this PR on behalf of Mike.
Fixes CODAGT-451
Adds optional `model_intent` metadata to the built-in execute tool
schema so tool calls can carry a short user-facing intent label without
duplicating the command or duration.
The Agents UI now composes that intent with the existing execute command
and duration fields, displaying labels like `Checking repository state
using git fetch origin for 2.3s` while keeping the shell command visible
as the audit-relevant action.
Existing execute calls without an intent keep the previous `Ran
<command>` fallback label, so only intent-bearing calls get the new
composed label.
Advisor tool calls currently reject questions over 2000 runes, which can
leave the parent model retrying the same invalid call.
This documents the limit in the advisor tool schema and guidance, then
truncates oversized questions rune-safely before building the nested
advisor prompt.
> Mux working on behalf of Mike.
Anthropic replay can fail when stored history contains a
provider-executed tool call like `web_search` without the matching
provider-executed result. That orphaned call is incomplete
provider-internal state, so replaying it can make an otherwise usable
chat unreplayable even though there is no search result to preserve.
This fixes replay by dropping orphan provider-executed tool calls from
the model-visible prompt, preserving signed reasoning and the rest of
the assistant content, then revalidating before the request. We do not
synthesize tool results or drop reasoning. The database can retain the
historical artifact for inspection, while Anthropic only sees replayable
content.
This matches permissively licensed prior art. Vercel AI SDK
(Apache-2.0), used by mux, keeps incomplete tool state in UI/history but
omits it from model requests with `convertToModelMessages(..., {
ignoreIncompleteToolCalls: true })`. LangChain, LiteLLM, and OpenAI
Agents (MIT for the relevant open-source code) also preserve Anthropic
signed reasoning as opaque replay data. Coder applies that model-visible
replay boundary explicitly because our persisted history is already in
provider-message form.
This matches mux, is cleaner than the older idea around not persisting
the search query tool, and the model handles the repaired prompt fine.
Closes CODAGT-448
## Before
<img width="963" height="491" alt="image"
src="https://github.com/user-attachments/assets/a7788ebf-2728-4420-90cf-5e4f6905bdf7"
/>
## After
<img width="842" height="513" alt="image"
src="https://github.com/user-attachments/assets/ae39c262-7586-4e2d-b7db-1b639a7e8e15"
/>
`attach_file` was registered for plan-mode turns but never added to
`builtinPlanToolAllowed`, so the per-turn `ActiveTools` allowlist
filtered it out and calls failed with `Tool not active in this turn:
attach_file`. This was an omission rather than a deliberate block — the
tool (#24280) landed shortly after plan mode (#24236) and no subsequent
edit to the allowlist picked it up.
Add `attach_file` under the `isRootChat` case, matching how other
artifact-producing tools (`propose_plan`, `write_file`, `edit_files`)
are gated. The tool only reads from the workspace and writes to
chat-attachment storage, so it preserves plan mode's invariant of not
making implementation changes to the workspace. Subagents in plan mode
remain restricted to the minimal read-only surface.
Records reasoning start and end times on persisted reasoning
`ChatMessagePart`s so reasoning duration can be computed for stored
chats. Backend-only: no SSE changes and no frontend rendering ship in
this PR.
The `created_at` field on `ChatMessagePart` is extended to also be
present on `reasoning` parts (it previously appeared only on `tool-call`
and `tool-result`), and a new `completed_at` field is added for
`reasoning` parts.
### How timestamps are recorded
- `StreamPartTypeReasoningStart`: stamp `startedAt = dbtime.Now()` on
the active reasoning state.
- `StreamPartTypeReasoningEnd`: stamp `completedAt = dbtime.Now()` and
append both into parallel `[]time.Time` slices on `stepResult`.
- Persistence reads the slices in occurrence order (reasoning has no
provider-side ID) and applies them to the matching `ChatMessagePart` via
`buildAssistantPartsForPersist`. The first reasoning block's stamps go
onto the first reasoning part, and so on.
- `flushActiveState` flushes partial reasoning interrupted before
`StreamPartTypeReasoningEnd` with `startedAt` from the active state and
`completedAt = dbtime.Now()` at the interruption.
### Why two fields, not one?
Tool calls and results are point events. The frontend computes their
duration by subtracting the call's `created_at` from the result's
`created_at`. Reasoning is one assistant part that brackets a span, so
we record both endpoints on the part itself.
### Why not stamp in `PartFromContent`?
Same rationale as #24101: `PartFromContent` is called during both SSE
publishing and persistence. Stamping there would yield incorrect
persistence-time timestamps for reasoning blocks that finished much
earlier in the step. Instead we capture in the chatloop and apply during
persistence.
<details><summary>Implementation plan</summary>
- `codersdk/chats.go`: extend `CreatedAt`'s `variants` to include
`reasoning?`; add `CompletedAt *time.Time` with `variants:"reasoning?"`.
- `coderd/x/chatd/chatloop/chatloop.go`: extend `reasoningState` with
`startedAt`; extend `stepResult` and `PersistedStep` with parallel
`[]time.Time` reasoning slices; stamp on
`ReasoningStart`/`ReasoningEnd`; thread the slices through all
`PersistStep` call sites including the interrupt-safe path; record
partial reasoning in `flushActiveState`.
- `coderd/x/chatd/attachments.go`: walk reasoning parts in occurrence
order and apply `step.ReasoningStartedAt[i]` to `part.CreatedAt` and
`step.ReasoningCompletedAt[i]` to `part.CompletedAt`.
### Tests
- `codersdk/chats_test.go` round-trips `created_at` + `completed_at` on
reasoning parts and verifies omission when absent and partial
interrupted parts.
- `coderd/x/chatd/chatprompt/chatprompt_test.go` asserts
`PartFromContent(ReasoningContent{})` does NOT stamp timestamps.
- `coderd/x/chatd/chatloop/chatloop_test.go`
`TestRun_ReasoningTimestamps` drives a stream with two reasoning blocks
and verifies parallel slices, monotonicity, ordering, non-zero values,
and content-block ordering.
`TestRun_InterruptedReasoningFlushesTimestamps` cancels mid-reasoning
and verifies `flushActiveState` records a non-zero pair.
- `coderd/x/chatd/attachments_test.go` covers
`buildAssistantPartsForPersist` for normal interleaved reasoning,
partial (zero `completed_at`), and missing slices.
</details>
> Generated by Coder Agents.
Co-authored-by: Coder Agent <agent@coder.com>
## Problem
Mid-turn workspace MCP discovery was broken when an agent was still
cold-starting. `PrepareTools` in `chatd.go` flipped
`workspaceMCPDiscovered = true` *before* calling
`discoverWorkspaceMCPTools`, so a failed discovery attempt permanently
blocked retries within the turn.
Customer-reported repro:
- New chat with no pre-selected workspace.
- LLM calls `create_workspace` mid-turn at `23:35:05`.
- `PrepareTools` fires, dials the agent with a 30s timeout, dial times
out at `23:38:15`, `discoverWorkspaceMCPTools` returns empty.
- Agent connects at `23:38:29`, 14 seconds later.
- `workspaceMCPDiscovered` was already true, so `PrepareTools` never
retried for the rest of the turn. MCP tools only appeared on the next
user message.
A naive retry loop in `PrepareTools` would also miss the bigger picture:
a workspace boot can take several minutes (EC2 cold start, 10 min
startup scripts), and the chatloop only gets a chance to call
`PrepareTools` between LLM steps.
## Fix
Do the workspace MCP discovery from inside the tool that already waits
for the agent. `chattool.CreateWorkspace` and `chattool.StartWorkspace`
call `waitForAgentReady`, which has a 2 min agent-online budget plus a
10 min startup-script budget. By the time they fire `OnChatUpdated`, the
agent is `Ready`. The chatd `onChatUpdated` callback now launches an
async `primeWorkspaceMCPCache` goroutine on every bind that has a valid
workspace ID:
- The primer calls `discoverWorkspaceMCPTools` until it returns a
non-empty list or `workspaceMCPPrimeMaxWait` (30s) elapses, with a 2s
backoff between attempts. The bounded wait handles the short race
between agent-online and the agent's MCP `Connect` settling.
- The primer runs asynchronously so the tool itself never blocks. Some
templates simply do not advertise MCP tools, in which case the primer
would otherwise spend its full budget for nothing.
- The primer shares the chat `ctx` (not a detached one) so it is
canceled together with the chat. A dangling primer would re-dial the
workspace conn after `runChat`'s deferred `workspaceCtx.close()` and
leak that conn.
- `inflight.Add(1)` ensures server shutdown still waits for any
in-progress primer.
- `PrepareTools` is simplified back to a single discovery call. It now
only sets `workspaceMCPDiscovered = true` on success, so an empty result
no longer permanently blocks discovery within the turn. The cache hit
warmed by the primer makes that call cheap in the common case; the dial
fallback handles the rare cache miss.
## Tests
All in `coderd/x/chatd/chatd_internal_test.go`:
- `TestPrimeWorkspaceMCPCache_SuccessOnFirstAttempt` — single
`ListMCPTools` call returning tools populates the cache.
- `TestPrimeWorkspaceMCPCache_RetriesUntilToolsAppear` — first call
empty, second returns tools; primer retries past the backoff and writes
the cache. Uses `quartz.Mock.Trap` on `NewTimer`.
- `TestPrimeWorkspaceMCPCache_GivesUpAfterDeadline` — `ListMCPTools`
always empty; primer stops at `workspaceMCPPrimeMaxWait` and refuses to
cache the empty result so PrepareTools can retry on the next step.
The existing integration test
`TestRunChat_WorkspaceMCPDiscoveryAfterMidTurnCreateWorkspace` continues
to pass and now also exercises the async-primer path end-to-end via the
create_workspace tool.
```
go test ./coderd/x/chatd/... -count=1
go test ./coderd/x/chatd/ -race -count=1
make pre-commit
```
<details>
<summary>Design notes</summary>
- The first iteration of this PR added retry+cooldown+failure-cap logic
inside `PrepareTools`. It worked for the customer's ~30s race window but
did not help workspaces that take several minutes to boot, because
`PrepareTools` only fires between LLM steps. Reviewer pointed out the
right place to handle this is the tool itself; the current
implementation does that.
- Why async: a primer that ran synchronously inside the `OnChatUpdated`
callback blocked the create_workspace tool from returning for up to
`workspaceMCPPrimeMaxWait`, which broke
`TestCreateWorkspaceTool_EndToEnd` and would hurt any template that does
not expose MCP tools. Decoupling lets the tool return immediately and
lets the primer warm the cache concurrently with the next LLM step.
- Why share the chat `ctx` rather than `context.WithoutCancel(ctx)` (the
title-generation pattern): the primer touches
`workspaceCtx.getWorkspaceConn`, which `runChat`'s deferred
`workspaceCtx.close()` invalidates. A detached primer outliving the chat
would dial a fresh conn and leak it.
- The constant naming distinguishes `workspaceMCPDiscoveryTimeout` (35s
per-call dial budget, unchanged from #25169) from
`workspaceMCPPrimeMaxWait` (30s total budget for the post-ready primer
loop) and `workspaceMCPPrimeRetryInterval` (2s between empty-result
retries).
</details>
Follow-up to #25169.
---
_This pull request was generated by Coder Agents._
Anthropic is strict about replaying the latest assistant turn once it
contains signed or redacted reasoning. We were still mutating that turn
in a few Coder-owned places: dropping empty reasoning blocks on replay,
rewriting provider-tool history during sanitization, and in the worst
case sending a prompt we already knew Anthropic would reject.
This patch keeps the latest signed assistant immutable through Coder's
replay and sanitization paths, preserves empty signed or redacted
reasoning anywhere Coder owns the ledger, and fails before the provider
call if the prompt is still unsafe.
It also bumps the existing `coder/fantasy` `coder_2_33` fork that `main`
already uses to the commit containing coder/fantasy#35. These fixes have
also been upstreamed to charmbracelet/fantasy.
Closes CODAGT-409.
> Mux updated this PR on behalf of Mike.
## Stack Context
This stack splits experimental personal skills into smaller reviewable
PRs. Personal skills are user-owned `SKILL.md` files stored by Coder and
injected into chatd alongside workspace skills.
Stack order:
1. #25362 personal skill resolver
2. #25363 storage, permissions, API, and SDK
3. #25365 API test coverage
4. #25366 chattool and chatd integration
5. #25066 settings UI and docs
6. #25386 personal skills slash menu
## What?
Adds the shared personal skill parser and resolver package, plus
reusable skill-name validation exported from `workspacesdk`.
The parser enforces the full personal skill contract: max raw size,
kebab-case name, max name length, and non-empty body.
## Why?
The rest of the stack needs one source-aware resolver for personal and
workspace skills, including collision handling and qualified aliases.
Keeping personal skill constraints in the parser prevents callers from
accidentally parsing invalid personal skills.
## Validation
- `go test ./coderd/x/skills ./codersdk/workspacesdk`
- pre-commit hooks on this branch
These chatd tests are flaking for the same stale control-notification
race tracked by CODAGT-353, so this change skips the newly reflaking
advisor-chain and `TestPatchChatMessage/ChangesModel` tests and rewrites
the older `TODO(hugodutka)` skips to point at the same root cause. This
keeps the known flakes documented consistently until the chatd
notification-flow refactor lands.
Closes CODAGT-427
Closes https://github.com/coder/internal/issues/1510
Mid-stream HTTP/2 peer resets from LLM providers can arrive after a 200
streaming response has already emitted provisional parts. Previously
those resets fell through as generic non-retryable errors because
`stream ID` messages did not match retryable transport signals, and
stream IDs could be misread as HTTP statuses.
Classify retryable HTTP/2 RST_STREAM codes as transient timeout
failures, ignore stream IDs during status extraction, and keep the
existing `retry` event as the rollback boundary for provisional message
parts so replacement attempts do not replay failed-attempt output.
Closes CODAGT-382
> Mux working on behalf of Mike.
## Summary
- retune chatd subagent guidance to prefer `general` for substantial
delegated work, including read-only synthesis and planning support
- narrow `explore` guidance to repository-local code lookup and bounded
tracing
- add regression tests for planning, spawn tool, and Plan Mode guidance
text
## Tests
- `go test ./coderd/x/chatd -run
'Test(DefaultSystemPromptPlanningGuidance_SteersSubagentSelection|SpawnAgent_DescriptionSteersGeneralForSubstantialResearch|SpawnAgent_PlanModeDescriptionOmitsComputerUse|PlanningOverlaySubagentGuidance_UsesPlanModeSafeDescriptions|ExploreSubagentIsReadOnly)$'`
- `make lint`
- `make test TEST_PACKAGES=./coderd/x/chatd RUN=Guidance && make test
TEST_PACKAGES=./coderd/x/chatd RUN=Description`
- pre-commit hook during `git commit`
Ethan