mirror of
https://github.com/coder/coder.git
synced 2026-11-02 05:23:10 +08:00
d014d4c372fb7dfdb3b0ea00a9935fec9b9528f9
6
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Michael Suchacz
|
48e1e28638 |
fix(coderd/x/chatd/chattool): make edit_files schema and errors actionable for models (#28121)
## Problem Chat `45b87e40-ffe7-49e5-8932-5fd0bdb9e542` on dev.coder.com failed 57 of 75 `edit_files` tool calls. Every failure was the same: the model omitted `files[].path` (it batched edits per file but only filled in `edits`), and the error relayed back to the model was: ``` POST http://[fd7a:115c:...]:4/api/v0/edit-files: unexpected status code 400: "path" is required ``` The model retried the identical malformed call dozens of times. Two gaps made this sticky: 1. The `edit_files` input schema had no field descriptions, so `path` was only a bare required property. 2. The agent API error reached the model wrapped in HTTP transport noise (method, internal tailnet URL, status code) with no indication of which `files` entry was broken. ## Changes - Add `description` tags to every `edit_files` schema field and state the path requirement in the tool description. - Validate `files` entries in the tool before plan-turn checks and the workspace connection lookup, returning entry-indexed errors such as `files[1].path is required; provide the absolute path of the file to edit; no files in this batch were applied`. - Relay agent API failures with `Message`, `Helper`, `Detail`, and `Validations` from `codersdk.Error` instead of the raw transport-prefixed string. ## Validation - `go test ./coderd/x/chatd/chattool` passes; new tests cover the schema description, entry-indexed validation errors, and transport-noise stripping (each verified red-green by toggling the fix off). - `go build ./...`, `go vet`, and pre-commit (fmt + lint) pass. > Mux created this PR on Mike's behalf. |
||
|
Michael Suchacz
|
c17bed25e0 |
feat: wire chat lifecycle hooks into chatd (#27429)
Wires chat lifecycle hooks into chatd, gated by the `agent-lifecycle-hooks` experiment. Part of the lifecycle hooks stack (#27401, #27428, #27430). See `docs/admin/setup/chat-lifecycle-hooks.md` for the consumer-facing contract. ## Summary When a hook URL is configured, chatd dispatches `session_start`, `user_prompt_submit`, `pre_tool_use`, `post_tool_use`, `pre_compact`, `post_compact`, and `stop` events to the consumer and applies its responses. ## Design - **Stateless**: Coder stores no hook dispatch or decision state. Delivery is at least once; consumers deduplicate on stable payload identifiers (chat ID, event type, tool-use ID) and answer duplicates with the same decision. - **Admission-time prompt effects**: `user_prompt_submit` dispatches exactly once per submission (create, send, queue, edit, subagent spawn) and folds its effects into the stored prompt as typed message parts: original-or-overridden user parts, then model-only `hook-context`, then a user-visible `hook-notice`. Hook context is stripped from every client-facing conversion; hook notices are excluded from model prompts. The server rejects hook parts in client-submitted content. - **Tool gating**: `pre_tool_use` allow can override tool input; deny becomes a synthetic denied tool result, with any returned model context persisted as a model-only transcript row so it never reaches clients. The denial text identifies an external policy (the deployment's lifecycle hook) as the source and marks the decision as persistent, so the model explains the denial instead of retrying it or misreporting it as an infrastructure failure. - **Fail closed**: a dispatch failure rejects the triggering request or moves the chat to the error state in the same transaction as the affected step, so a runnable state is never published with unapproved content. - **Admission before persistence**: `pre_tool_use` is dispatched for the calls the model produced, before the assistant message is stored. See "Staged tool admission" below. - **Fresh dispatch per tool call**: every non-provider-executed tool call is decided by its own `pre_tool_use` dispatch; Coder never reuses an earlier decision on the consumer's behalf. Retries re-dispatch the same logical event. ## Structure All hook dispatch flows through one seam: entry points build a `chathooks.Chat` (chat identity) and a `chathooks.Message` (event details) and call `Trigger.Trigger`, the only component that talks to the dispatcher. The integration lives in the `coderd/x/chatd/chathooks` subpackage, split by responsibility: - `trigger.go`: the trigger seam; builds the wire envelope per event, normalizes deny into a typed error, and holds the package's single enabled-check. - `effects.go`: pure conversion of hook results into transcript rows and prompt parts. - `errors.go`: failure classification (dispatch error messages, denial mapping, tool-result dispatch-failure scanning). - `tooluse.go`: the tool-call gate (`pre_tool_use` preflight, `post_tool_use` payloads, applying admitted input to the step). Server-bound glue stays in `coderd/x/chatd/hook_server.go`: the chat-parking dispatch error handlers, the step-commit row insertion wrappers, and the dynamic post-tool-use state loader, which depends on chatd validation types. This PR adopts the `codersdk/x/agenthooks` and `coderd/x/agenthooks/dispatch` import paths introduced at the tip of #27401; intermediate commits still reference the pre-move paths and are not individually buildable. ## Staged tool admission `pre_tool_use` originally ran at tool execution time, which is after the assistant message carrying the tool call was already committed. An `input_override` therefore had to rewrite stored message content in place. @hugodutka pointed out that chatd treats message content as immutable, and that the rewrite was a shortcut rather than a requirement. It was also a correctness problem in its own right: the rewrite only updated the database, so the transcript could show one input while a different one had executed. The hook now runs before the step is persisted: ```text provider stream ends (tool calls complete, in memory) -> pre_tool_use dispatch per call -> ONE transaction: assistant row with admitted inputs, synthetic denials, hook rows -> execute ``` The step is inserted once, carrying the input the tool runs with. `UpdateChatMessageContentByID` and `Tx.UpdateMessageContent` are deleted from #27428, so message content stays immutable. Two consequences, both intentional: - **Clients converge rather than wait.** Tool-call parts still stream live, so a rewritten call briefly shows the model's proposed input before the committed message replaces it. The chat store already clears stream state when an assistant message arrives, so the stored input wins with no frontend change and no added latency before tool cards appear. - **A call already in history was already admitted.** Execution consumes the stored input instead of dispatching a second decision, which keeps one dispatch and one set of hook effects per call. A consumer policy change between admission and execution applies to later calls, not to calls already admitted. The per-chat debug endpoint still records the provider's original tool input. Its purpose is to report provider behavior, and it requires an explicit per-chat debug flag; the invariant here covers the transcript. ## Configuration Adds `chat-hook-url`, `chat-hook-secret`, `chat-hook-timeout`, and `chat-hook-enabled` deployment options with startup validation. The flags are hidden from `coder server --help` while the feature is experimental; the setup guide documents them. ## Tool input validation Built-in tool arguments reach a consumer as raw JSON with key spelling preserved, but the tools decode those bytes with Go, which matches struct fields case-insensitively and keeps the last match. A policy reading `path` could therefore authorize one value while the tool executed another, and a lone case variant such as `{"PATH":"/secret"}` was invisible to a policy checking for `path`. Coder now rejects a built-in tool call whose input repeats a key or spells a schema property with different capitalization, before the `pre_tool_use` dispatch, so a consumer is never asked to authorize bytes whose meaning depends on the reader. Rejected calls produce an error result the model can retry; unambiguous calls in the same batch still run. A consumer-authored `input_override` is rechecked after the dispatch and fails the turn closed, because the model cannot correct it. Dynamic and MCP inputs are excluded because the client and the workspace agent execute those calls rather than coderd. Two paths needed more than a schema check. Execution resolves a deprecated tool name to its canonical tool, so validation resolves aliases first. The `edit_files` decoder also reads `search` and `replace`, which its schema does not advertise, so those aliases are now matched exactly and their case variants ignored. A hook denial now returns a structured 403 carrying `kind: "hook_denied"`, mirroring the dispatch-failure response that already carries its own kind. Without it a client cannot tell a policy decision apart from a generic failure, and the chat UI titled a denial "Request failed". Adding a kind needs no migration: `ChatErrorKind` is persisted only inside the JSONB `chats.last_error` column, whose decoder accepts unknown kinds. The hook docs also correct the tool-input convergence window. A batch dispatches sequentially before the assistant row commits, so the original input stays visible for a span that scales with the number of tool calls in the step rather than a single hook timeout. > This PR was written by Mux, an AI coding agent, on Mike's behalf. |
||
|
Mathias Fredriksson
|
32ed9f1f39 |
fix: use old_text/new_text in edit_files tool schema (#25658)
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.
|
||
|
Mathias Fredriksson
|
6b0bb02e5d |
fix: server-side diffs and stricter fuzzy splicing for edit_files (#24454)
Fixes three classes of edit_files bugs and adds structured per-file
diff output for tool callers:
- New IncludeDiff flag on FileEditRequest; when set, the agent
returns FileEditResponse.Files[]{Path, Diff} with unified diffs
computed via go-udiff v0.4.1 Lines + ToUnified (not Unified,
which calls log.Fatalf on internal error).
- Fuzzy match comparators split each line into leading whitespace,
body, trailing whitespace, and ending. The splice substitutes at
each position: on agreement between search and replace the file's
bytes win; on disagreement the replacement's bytes are spliced
verbatim. Carve-outs for empty-body lines, multi-line EOF splices,
and level-aware indent translation for inserted lines.
- Indent-unit detection (GCD for spaces, tab-priority) lets a 4sp
LLM search insert correctly into tab or 2sp files. Falls back to
the previous cLead-inheritance path when units can't be detected
cleanly.
- Empty search is rejected with "search string must not be empty".
- Duplicate file paths in one request are rejected; symlink aliases
resolved via api.resolvePath before the dedup check.
- Frontend EditFilesRenderer consumes the structured files array by
explicit path (no label munging) with per-file synthetic fallback
for older agents or mismatched paths. On error, no diff is
rendered so the synthetic fallback doesn't misrepresent a
rejected edit as applied.
Breaking change: AgentConn.EditFiles changes from (ctx, req) error
to (ctx, req) (FileEditResponse, error) in codersdk/workspacesdk.
Source-breaking for external Go consumers; no compat shim per plan
owner.
Out of scope (tracked in CODAGT-214): level-aware indent for
middle-substituted splice lines. Locked in
TestEditFiles_FuzzyIndent_InsertionLevelAware's Lock_* cases plus
TestEditFiles_ReplaceAll_FuzzyIndentGap.
|
||
|
Michael Suchacz
|
1cf0354f72 |
feat: add plan mode with restricted tool boundary (#24236)
> This PR was authored by Mux on behalf of Mike. ## Summary - add persistent plan mode for chats and the chat-specific plan file flow - add structured planning tools such as `ask_user_question` and `propose_plan` - keep `write_file` and `edit_files` constrained to the chat-specific plan file during plan turns - allow shell exploration in plan mode, including subagents, via `execute` and `process_output` - block implementation-oriented, provider-native, MCP, dynamic, and computer-use tools during plan turns - update the chat UI, tests, and docs for the new planning flow |
||
|
Michael Suchacz
|
a554de372a |
fix: use per-chat plan file paths (#24268)
> This PR was authored by Mux on behalf of Mike. Chats sharing one workspace (e.g. sibling subagents) all wrote to `/home/coder/PLAN.md`, causing plan file collisions. This change derives a unique plan path per chat from the workspace home directory and chat ID. ## Changes * `write_file`, `edit_files`, and `propose_plan` reject any `plan.md` variant (case-insensitive) at the workspace home root, with a clear error pointing to the chat-specific path. * Root chats receive a `<plan-file-path>` block inlined in the main system prompt with the concrete path. * Prompt and tool descriptions no longer hardcode `/home/coder/PLAN.md`. * Plan path handling is POSIX-only (forward-slash), relying on the contract that workspace agent paths are normalized before reaching chatd. * Updated `ProposePlanTool.stories.tsx` to use per-chat path examples. * Full test coverage for plan path detection, legacy-path rejection in all three tools, inline prompt rendering, and fallback behavior. |