Add a server-pushed run_event frame so a client no longer has to poll
once a second to notice run changes.
- protocol: RunEventFrame/RunStepDelta compact delta (run id, workspace,
status, outcome, current stage, completed-step count, timestamps and an
optional single step delta), plus subscribe/unsubscribe/run_event types
and the optional RunEvent/RunIDs message fields. All fields optional so
older peers ignore them.
- server: runEventHub with per-connection subscriptions scoped to a
workspace (and optional explicit run ids), one bounded writer goroutine
per subscription, slow-subscriber drop, per-client subscription cap and
a ~220ms per-run coalescer that flushes pending deltas before a terminal
frame.
- runner: emit deltas on created/queued, queued->running and step start,
human waiting, step finish (outcome, target, diagnostic, artifacts and
completed count), terminal finish, cancel of a waiting run and
decide/resume; emitRun is a no-op without a hub.
- websocket: subscribe seeds current state before the ack, unsubscribe and
connection teardown drop subscriptions, and pushed frames reuse the
connection write closure.
- tests: protocol golden-JSON wire shape plus websocket coverage for the
started-run frame sequence, subscribe current-state snapshot, silent
unsubscribed/never-subscribed clients, workspace isolation and a
bounded coalescing frame count.
- docs: note subscribe/unsubscribe/run_event in docs/cli.md; TUI and
Android still poll in this release.
Protocol: replace the loose Query/Kind/Matches/Limit search fields with a Search request object and a SearchResults response, and have the server answer search from Store.Search over the shared run projections (runFrom/runStamps).
CLI: add the read-only 'pipelines search QUERY [--type KIND] [--limit N] [--json]' command with usage/help, and document the typed targets, ranking and bounds in docs/cli.md.
TUI: the '/' popup keeps type-to-search, up/down, Enter and Esc, renders one ranked list with kind badges and a bracketed (non-colour) match snippet, and Enter opens the matched pipeline, run, step, artifact or field; footer text updated to match.
TUI: add a dedicated run artifacts view ('A') with open/copy of recorded files, and move artifacts out of the run-detail File changes section (artifactCaptureNotes).
Tests: add Store.OpenReadOnly for read-only inspection of a live database plus pipeline/server/TUI search tests and artifacts tests, including env-gated live checks.
Add `pipelines run --model PROVIDER/MODEL` and repeatable
`--stage-model STAGE=PROVIDER/MODEL`, resolved with precedence
per-stage > whole-run > the model pinned in the pipeline version.
- A shared pure resolver (`pipeline.ParseModelOverride` /
`pipeline.EffectiveModels`) owns the precedence and validation: every value
must be a `provider/model` pair (`stage.ValidModel`), every stage key must
exist in the pinned graph, an unknown key fails naming the valid keys, and a
per-stage key that names a non-model stage is refused.
- The server validates against the pinned version before admission (no run row
is ever inserted for an invalid request), persists `model`/`stage_models` on
the run and `provider`/`model` on each step, applies the effective pair to a
clone of the stage definition so the stage and its repair/formatter path share
it, overlays the returned run graph with the effective pair (presentation
only), and reuses the recorded overrides on retry.
- The run record echoes the overrides it was launched with and each step records
the provider/model it actually ran with; the TUI Evidence pane renders that
effective pair. The pinned pipeline version and `show --version N` are never
modified.
- CLI help, `docs/cli.md` and the packaged authoring skill references document
the flags and the precedence. A TUI override picker remains a documented
follow-up; the CLI is the deliverable.
Tests: pure precedence/validation table tests, store round-trip and
request-key-mismatch plus migration, server admission/application/overlay/retry
with a controlled executor, CLI flag/help/malformed-value tests, and a TUI
evidence test.
Introduce a short, markdown-ish result format whose first non-empty line is
the declared outcome and whose "# output" section carries the routing data,
so stage output stops carrying multi-kilobyte prose inside a JSON envelope.
- New internal/stage/line_result.go: resultFormat() gates legacy JSON ('{'
first) vs the new line format; decodeLineResult() parses outcome-from-first
line, requires a JSON-object "# output" section, stores extra sections
verbatim under normalised keys, and records a reserved result_parse status
(parsed/partial/failed). Failures return a salvage Result plus
ErrInvalidModelResult, and never route an undeclared outcome.
- internal/stage/model.go: write the raw response to raw-output.md (0600,
best-effort) before parsing, and route the line format through
decodeLineResult while leaving the JSON path (stageResultText ->
decodeModelResult -> repairModelResult) unchanged.
- internal/stage/agent.go: agentResultText returns line-format text verbatim
so the "# output" body is not stripped by the JSON-object extractor.
- internal/server/runner.go: merge a non-empty salvage Result into the error
envelope under "salvage" on failure (a no-op for the legacy JSON path).
- internal/stage/line_result_test.go: raw-text fixtures covering happy path,
missing/undeclared outcome, missing "# output", best-effort extras,
adversarial prose, legacy JSON unchanged, >=64KB responses, and
raw-output.md written before parsing.
Exchange a build/protocol identity so an older CLI can no longer talk to a
newer native server silently:
- protocol carries ProtocolVersion and every server frame (including the
unsolicited welcome) is stamped with the server build and protocol, so a
client that never pings still learns the server identity.
- buildinfo.Compare classifies the relationship as InSync/Warn/Refuse and
names both builds; the client Watch reports one warning per connection and
returns a hard error when the server speaks a newer, uninterpretable
protocol so the caller never acts on the frame.
- client.ReceiveNotice decodes the raw frame so an unmodeled top-level field
produces a one-time "unrecognized field(s)" notice instead of being dropped
silently (the case that hid the artifacts field).
- CLI routes warnings to stderr, keeps machine JSON stdout clean and fails
non-zero on a refused mismatch; ping/connection print the client/server
relationship. The TUI surfaces the same notices through the shared log.
- scripts/setup-native.py now warns when an older `pipelines` earlier on PATH
shadows prefix/bin/pipelines, naming both paths and versions, without
installing or modifying anything else.
- Docs, the packaged debugging skill reference and README no longer imply the
installed CLI always matches the server.
Tests cover the comparison table, unknown-field detection and caching, the
server identity stamping on pong and welcome, the CLI/TUI reporting paths and
the installer PATH shadowing warning.
A new registered stage type `human` (internal/stage/human.go) has no model,
agent, tools or configs: its required `prompt` block carries the operator
instructions and its non-error edges are the selectable options. It parses,
validates, renders, exports and imports like every other type (no parser
special-case, no config keys).
At runtime the runner parks the run instead of executing: the step and the run
persist `waiting` with self-contained {prompt, options} evidence, and the run
returns without starting later stages, capturing file diffs, scanning artifacts,
holding a goroutine/timer or applying a stage timeout. The wait is durable: a
restart leaves the waiting rows untouched (RecoverRuns excludes them) and a
rebuilt runtime over the same database still accepts the decision.
`pipelines decide <run-id> [stage] --option KEY [--feedback TEXT] [--json]` and
the TUI `c` picker resolve the decision from the pinned graph in one store
transaction, record the choice and feedback as the step's evidence and resume
along edges[KEY]. An unknown option is refused with the valid list, a stage that
is not waiting is refused, and a repeated decision for an already-decided visit
does not double-apply. Cancel and retry keep working while waiting and after a
decision; a human stage is never killed by a stage timeout or max_visits, and a
`notify` policy can fire an `on=waiting` event.
Docs: the authoring skill references (format.md, operations.md), docs/cli.md and
README describe the stage, the command and the TUI picker. Tests cover stage
validation, the parsing round-trip, the store transitions, a websocket wait/decide
integration, cancel/retry, waiting notifications and a persistence-layer rebuild
(second Open + newRunner over one database, no process restart).
Each completed step now exposes a verdict: the chosen option (outcome -> target),
a bounded one-line reason with model/fallback/absent provenance, and the recorded
artifacts the reason refers to. The store persists only the model-generated line
in a new step_verdicts table and derives the fallback at read time (no backfill);
the server generates the line once at completion, best-effort; the TUI renders it
in the stage progress summary. The packaged debugging reference documents it.
Add a bounded, never-empty title to every run. `pipelines run --title "..."`
sets it explicitly and skips model generation; otherwise the server generates
it from the prompt through the shared summariser and falls back to the prompt's
first line, recording title_source (model/fallback/explicit) and title_model.
- protocol/store: RunSummary.Title plus Label(), a set_run_title message, title
columns with an idempotent migration, CreateRunTitled, and SetRunTitle renames.
- summarise: Title/GenerateTitle/DeriveTitle/ExplicitTitle bounded to TitleCap.
- runner: generate the title outside the admission lock and only for a request
that starts a run; a replay reuses the stored title and an explicit title
skips the model entirely.
- CLI: `--title` on run and a `title <run-id> "<title>"` rename command; human
`runs` and `inspect` show the title beside the pipeline name and version.
- TUI: the run list row, run detail header, Home run cards and last-run line,
and the pipeline Results line show the title via Label().
Tests cover storage/migration/fallback, explicit-title and rename, the
end-to-end CLI flow, and title rendering; CLI and operations docs updated.
chat_sessions now brokers GET /api/session but filters the global list to
the sessions this server recorded in run_steps, deduped by session id, and
fails closed to an empty list when no ownership source is configured. A
session another client holds (in-server claim, released on stream end or
disconnect) or an active run is marked busy; a second client's chat_prompt
for a held session is rejected before any OpenCode call. chat_create stays
unbrokered.
The chat_interrupt path was already wired end to end (protocol type,
server dispatch through handleChatSimple to broker.Interrupt, and the
adapter POST /api/session/{id}/interrupt), but nothing exercised it
against a live stream. Add the missing coverage:
- internal/opencode/chat_test.go: pin the adapter's outbound call
exactly (POST /api/session/ses_1/interrupt, empty JSON body, one
request, nil on 200) plus a non-nil error on HTTP 500.
- internal/server/chat_test.go: drive the real opencode.Chat adapter
over an httptest SSE fixture that holds the turn open until the
interrupt arrives, then assert the run ends, the interrupt is
acknowledged, no error frame is sent, and the prompt/interrupt posts
and interrupt path match. Also cover invalid-session rejection,
the unconfigured-broker error, and broker-failure reporting, adding a
spare interruptErr hook to the scripted fakeChat.
Production code is unchanged: verification found no gap in the existing
interrupt transport. Tests use httptest loopback fixtures only and do
not touch the live OpenCode service.
decodeChatMessages only understood the native-serve {info,parts} envelope,
so records from the V2 HTTP API (GET /api/session/{id}/message) decoded
with an empty role and were silently dropped, leaving chat_messages with
empty history. Teach nativeChatMessage to read the V2 top-level shape
(id/type/agent/model/time, parts under "content", user prompt as a bare
string) while keeping the native {info,parts} path via role/id/model/agent/
created accessors and a contentParts fallback.
Add seeded-session evidence: a decoder table test covering both envelopes
(user + assistant with text, reasoning and a completed tool part carrying
input, content and status) plus handler tests that drive chat_messages
through the broker fallthrough and through the real adapter.
Add the server-side half of the brokered chat path (route B): the
opencode.Chat adapter forwards chat_prompt to POST
/api/session/{id}/prompt through the owned Connection.Call translation
and derives chat_event frames from the authenticated /api/event SSE
firehose (text and reasoning deltas, tool parts with their status
transitions, step and usage). Every pushed frame carries a stable
session and message id, with a last-message-id fallback so step/tool
events that omit the assistant message id stay attributable.
The /connect handler keeps per-session retained state (chatHub) capped at
1 MiB with least-recently-updated eviction at 128 turns, so a
reconnecting client re-syncs through chat_messages instead of replaying
deltas the OpenCode per-session log never resends. Prompt dispatch
subscribes before posting and buffers pre-ack frames to preserve order,
and cancels the stream on disconnect so no goroutine leaks. chat_sessions
and chat_create are declared but not yet brokered: they answer with an
explicit protocol error rather than the generic Pong.
Reuse the existing transport, validation and normalization; no second
OpenCode transport or credential path is introduced.
Tests: protocol golden round-trip, adapter normalization + v1/v2 Call
translation, and handler ack/stream/validation/resume/disconnect, plus a
real-adapter no-tool evidence harness.
Fix four stage failure-handling and observability defects (A2, A3, A6, A9).
- A2: add safeToReplay(def) as the single definition of retryable post-admission
work - only an explicit tools block that selects neither edit nor shell. The
agent/pi WorkStarted guards now break only when the stage is not safe to
replay, so read-only stages genuinely retry a post-admission timeout while
mutating and legacy-bundle stages still never replay admitted work.
- A6: re-emit a malformed final result on an explicit read-only stage up to the
retry budget, preserving format_repairs evidence; mutating stages keep the
immediate return and their session.
- A3: timed-out stages retain usable evidence. waitForSessionIdle records one
bounded sanitized OperationEvidence on deadline; the agent deferred cleanup
does a bounded best-effort latest-assistant readback and settles knowable V2
usage; RetainSession adds a retained_session pointer to failed step
transcripts so the session is visible beneath the BoundTranscript threshold.
- A9: a stage outcome routed to #failure now yields a non-empty run diagnostic
("stage <name> returned outcome <outcome>") while preferring an existing step
diagnostic.
Adds focused stage and server tests for retry safety, malformed-result
re-emission, timeout evidence retention and run-level diagnostics; updates the
packaged skill references and docs/cli.md, docs/model-results.md.
Every agent stage now writes into a per-run/step artifacts directory under
the data dir (<data>/artifacts/<run>/<step>), passed to Pi as
PIPELINES_ARTIFACTS_DIR, exposed on the OpenCode client env seam, and stated
in the agent prompt boilerplate so agents save images, screenshots, reports
and logs there.
After each step the server scans that directory with a bounded, symlink- and
traversal-safe inventory (os.OpenRoot + Lstat + O_NOFOLLOW) recording name,
size_bytes, content type (extension then 512-byte sniff fallback) and URL,
enforcing max files, per-file bytes and total bytes while counting skipped
and truncated entries. A missing, unreadable or oversized directory yields
only a best-effort diagnostic and never changes the step outcome.
The list is carried on protocol.RunStep as Artifacts []StageArtifact plus
skipped/truncated counters (omitempty, backwards compatible), persisted in a
new step_artifacts table keyed (run_id, sequence) alongside the
notifications/file-changes tables and loaded by GetRun.
GET /artifacts/{run}/{step}/{name} on the existing handler serves the bytes
with http.ServeContent, so Content-Type, Content-Length, HEAD and Range
(206/416) come from the stdlib; unknown runs/steps, absolute paths,
traversal, symlink escapes and non-regular files are refused.
The TUI lists each selected step's artifacts (name + human size) in the
retained-evidence section and adds an artifacts summary in the style of the
file-changes summary. Adds docs/stage-artifacts.md linked from README and an
artifacts subsection to the packaged operations skill reference.
Tests cover the scan records and caps, missing/oversized directories,
HTTP HEAD/Range/traversal/symlink/unknown refusals and protocol round-trips.
The per-workspace concurrency added in 2618f0d exposed three places that assumed a
single run, and the second concurrent run failed with 'agent quiescence not
confirmed' because of them.
- ActiveAgentSessionsForRun(runID) scopes the end-of-step quiescence check to the
run's own sessions, so another run's legitimately active agent no longer fails
this one. A run still fails when its own session cannot be confirmed idle.
- The startup guard now uses ActiveSessionOwners (which carries the owning run id)
and refuses only an ORPHANED session - one whose run is not active now - which is
what restart recovery is actually about. A session owned by a currently active
run no longer blocks new work.
- sessionActive examines only the target session: foreign entries in the
/api/session/active map are ignored, the target's absence is inactive, and the
target present with an unsupported state still fails closed.
Verified: go build, go vet and go test ./... pass, including ten concurrency tests
(two workspaces run concurrently, same-workspace refusal, symlink and default-cwd
equivalence, idempotent repeats, restart recovery, retry scoping, overlapping runs
not failing quiescence, and a run's own active session still failing it).
Operator verification, which a stage cannot perform because restarting the server
would abort the run doing it: the service was rebuilt and restarted on this
commit, then two runs in separate workspaces (/tmp/ws-c and /tmp/ws-d) executed
concurrently - both reported executing at the same time, a third run in ws-c was
refused naming run 46, and both finished succeeded with no diagnostic.
Track each active run's resolved workspace alongside its cancel function
in a single runner.active struct map. A new request key is now refused
only when an already-active run is changing the same resolved workspace;
the refusal names the conflicting run id and advises waiting or using a
different workspace. Workspace resolution is unchanged (default cwd,
absolute request workspace, EvalSymlinks), so two spellings of one
directory still compare equal. The conflict check and registration stay
inside the existing r.mu critical section, so concurrent starts for one
workspace cannot both pass. Idempotent repeats of an active request key
and the restart-recovery guard are unchanged.
Add concurrency_test.go covering concurrent different-workspace runs,
same-workspace refusal, symlink/default-cwd equivalence, idempotent
repeats, restart recovery and retry workspace scoping; strengthen the
runner_test.go refusal assertion. Update docs/cli.md,
docs/first-version-testing.md and the packaged operations skill reference
to describe per-workspace concurrency.
Assisted-by: DeepSeek V4.1 Flash
Install current verified harness releases inside Docker, preserve persistent accounts and settings, and provide separate interactive and pipeline model defaults with terminal onboarding and fuzzy pickers.
Add Docker-aware authoring transport, Pi stage execution, bounded workflow controls, reliable final-result readback and usage settlement, and finite short-lived CLI pipe draining. Keep dated installation evidence and packaged authoring guidance consistent.
Verified with a real Docker installation, both harness stages and interactive authors, the built-in plan/implement/review starter, independent host race/vet checks, installer tests, and packaged skill tests. Generated release receipts stay ignored in clean checkouts.
Make OpenCode authoring responsive and expand the managed skill references. Add conflict-aware import/export, real-provider reference workflows, and tracked V2 generation for reliable budget accounting.