d76bf2be67addcdd1c39536cb8ffcfd1cea4792e
36
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
ed1557c12e |
feat(edastore)!: move event capture and bindings onto PocketBase collections
The event-capture tables (captures, mappings, bindings, binding_dispatches,
playbook_dispatches) now live in PocketBase-owned collections and their ids
are opaque record ids rather than autoincrement integers. The task lifecycle
stays on SQLite: tasks/transitions/events carry FIFO claim ordering, id-cursor
pagination and the task-grant map, none of which gain anything from being
hand-editable. binding_dispatches.task_id is the single seam and stays an
integer addressing that store.
BREAKING: CapturedEvent.Id, Mapping.Id, Binding.Id and Binding.MappingId are
strings on the wire, as are the ids in the capture/mapping/binding requests.
RecordDispatchRequest.task_id is deliberately unchanged.
What the collections buy, beyond storage: a payload nobody has modelled yet is
inspectable, a mapping is editable at runtime, and a binding's relation to its
mapping is enforced. The old schema declared no FOREIGN KEY on that link, so a
dangling mapping_id was previously possible and is now refused.
Behaviour preserved deliberately, each recovered from the SQLite version
rather than reimplemented from memory:
- draft -> pending_approval -> armed, with approval only from
pending_approval and an edit always returning a binding for re-approval
- one binding per source, guarded on insert, update and approve
- an empty secret on update keeps the stored one, so an edit form that does
not echo it back cannot silently disarm a binding
- ErrMappingNotFound / ErrBindingNotFound on a missing delete target
- sentinels are re-exported from storecontract, never redefined: staterpc
matches on exact canonical message to rehydrate them, so a second
definition would break errors.Is across the wire invisibly
Two implementation notes worth keeping:
- at-most-once dispatch is now a unique index rather than INSERT OR IGNORE.
Only the unique-violation maps to ErrAlreadyDispatched; any other save
error is returned, because treating them all as "already recorded" would
turn the ledger into a silent at-least-once dispatcher.
- captures store received_at themselves at nanosecond precision in a
fixed-width layout. PocketBase autodates are millisecond and record ids
are random, so same-millisecond captures would otherwise order
arbitrarily; the old store got insertion order free from autoincrement.
- ListUndispatchedCaptures is raw SQL: the anti-join is a subquery and
PocketBase's filter DSL is not SQL. A read fires no hooks either way.
Adds tool_calls, which had no persistence at all: tool calls only ever reached
the on-disk task log, readable as text but never queryable or watchable.
Nothing writes it yet.
The binding secret cipher moves with the bindings it protects.
|
||
|
|
b54f01065c |
feat(tools): add per-server parallel_tool_calls to the MCP client
MCPServer gains ParallelToolCalls (default false). The client's per-server callMu is now nil for a server that opted in, so CallTool stops serializing that server's calls and the caller's own concurrency decides; every other server keeps exactly one tools/call in flight. The flag is threaded from config through the provider constructor - the only place MCPServer configs become Client instances - into NewClient, in both the daemon and the agent worker. |
||
|
|
e1c4914873 |
fix(workflow): make gate.diff-rules fail closed on a change it never read
The step's contract said it diffs "the worktree". It does not: Trees.Diff reports the merge base against HEAD, so only committed work is visible. A stored definition that placed gate.diff-rules before the step that commits therefore ran every rule against an empty diff, found nothing and returned nil, letting the run proceed -- the "load and check nothing" failure the same file's settings validation exists to prevent, and indistinguishable in the run's events from a change that passed. The rules now read the committed change, and a run that read no committed change while the worktree still holds uncommitted work fails on the spot, naming the step it must follow. The count of added lines read is logged on every run, so "checked nothing" is visible rather than silent. Telling those two apart needs a fact Trees does not carry, so it arrives as the same shape of optional capability changeStatsReader is: an unexported interface in the domain, implemented where git is actually read (worktree.Manager.HasUncommittedChanges, which excludes ignored files the way CommitAll's own status check does) and forwarded by hybridTrees, the Trees every production run uses. A Trees implementation that cannot answer is refused rather than assumed clean: for a gate, silence about the change is a failure, not a pass. The four remaining review findings are recorded, not fixed, in archie-core-ze7y (the escalated .archie/stages/*.go decision, P1, which now blocks archie-core-8szi), archie-core-4kbg (discoverability), archie-core-s1jz (hunk-header and threshold hardening) and archie-core-vof9 (retry keeps the pinned definition, so a rule cannot be unblocked by editing it). |
||
|
|
1db8c8410e |
fix(archied): register the step vocabulary only where a root resolves it
Review pass on the step-vocabulary shape. The gateway root shared openStateStoreAdapter with the daemon, so it registered a step vocabulary and built a workflow-definitions client it never read: the same dead vocabulary the previous commit removed from internal/app/archiemessaging, reintroduced through the shared adapter, and able to fail the gateway's boot the day a provider set can fail to load. The vocabulary and the client now belong to the daemon root alone (openDaemonWorkflowDefinitions, called by Run before buildDaemon captures it), which is what Client's "a process that never reads or writes a workflow definition needs no provider set and cannot be failed by one" already claimed. The root guard was weaker than its own comment: it asserted that a root calls stepVocabulary(), not that the manager returned is the one handed to NewServer/NewWorkflowDefinitionsClient, so a root could call it and inject an empty manager and keep every test green. The pin now asserts the data flow (both halves mutation-checked), plus Run's ordering -- after openStateStoreAdapter, before buildDaemon. Three root guards asserted exact equality with workflow.BuiltinStepRegistry(), which contradicted the extension path they guard: adding a provider to workflowsteps.Providers() -- the documented way a step type arrives -- failed all three, and the expectation could not be satisfied by editing it. They now assert what their doc says (every shipped stage present) plus provenance (every name resolved is declared by the shared provider set), and the workflowsteps guard derives its expectation from Providers() so both sides move together. New guard on the defect class the bead exists for: TestNoProductionSiteReaches- ForTheBuiltinStepRegistry scans the module and fails if any non-test file outside the domain declaration and the shipped provider names BuiltinStepRegistry, which the agreement test cannot see for a sixth resolution site added later. Claim-vs-code corrections: openStateStoreAdapter "reads and replaces" -> the daemon only reads, and ReplaceWorkflowDefinitions has no production caller yet; workflowsteps' package/Providers docs no longer claim every Archie binary registers the set (archie-gateway, archie-messaging, archie-playbooks and archie-ui do not); the shipped-provider test no longer says production does not call BuiltinStepRegistry, which is false -- production calls it only there. |
||
|
|
67662b12cb |
fix(workflow): pin the roots, drop the dead vocabulary, guard the nil manager
Third review pass on the step-vocabulary shape, one fix per finding. P1, roots unpinned. The agreement test built its own manager and injected it, so it pinned the resolution sites and not the roots that decide the provider set; a root regressed to workflow.NewManager() (no shipped stages) left go test green. archied now resolves through one unit, stepVocabulary (step_vocabulary.go), and agentworker's productionWorkerDependencies is exercised directly: internal/app/agentworker/step_vocabulary_test.go and internal/app/archied/step_vocabulary_test.go assert each root's manager resolves the shared provider set's shipped stages and accepts a provider registered through it. RunStateStore and openStateStoreAdapter own a SQLite file and a gRPC dial, so their calls are pinned by source position, the way composition_order_test.go pins the memory-engine ordering. P2-B, dead vocabulary. archie-messaging registered a step vocabulary it never resolved (its only control-plane call is RuntimeChatConfig) and could fail startup on it before the State Store target guard. The workflow-definitions surface is now its own type, controlplane.WorkflowDefinitionsClient, so the vocabulary is a constructor dependency of the one surface that resolves step types and of nothing else. The messaging root no longer imports workflowsteps. P2-C, nil manager. A nil *workflow.Manager dereferenced on every resolution path. stepRegistry and CompilePinnedWorkflow now refuse with an error naming the wiring mistake; TestNilStepVocabularyFailsClosedAtEveryResolutionSite pins the refusal. Claim-vs-code corrections: agreement is stated as within one build (vocabulary.go, stepRegistry, openStateStoreControlPlane, openStateStoreAdapter); the Yaegi claim now says no bridge to StepFactory has been built and cites internal/secret/secretextract; the provider set names its three roots; Manager.Register's ordering is documented as a constraint with its consequence, and the shape rule no longer claims to reserve shipped stage names; the shipped-stage test compares against the builtin factory's own stage. |
||
|
|
005179ebf3 |
feat(workflow): own the step vocabulary in a family manager registered at each composition root
archie-core-fwmp. Every call site hardcoded workflow.BuiltinStepRegistry(), so the step vocabulary a workflow definition may name was closed to the six shipped workflows and archie-agent could compile only what the control plane happened to validate against the same literal. internal/domain/workflow.Manager is the family owner (ARCHITECTURE.md#plugin- engine-rule-strict invariant 2): a constructed, injected value with no package-level registry and no init()-time registration. Register validates at the producer and atomically -- nil or unnamed provider, malformed provider or step-type identifier, a step type with no factory, a duplicate claim, a provider declaring one type twice, and a contribution shadowing a shipped stage are each refused, and a refused contribution applies nothing. Registry() hands back a copy, so the closed set is a property of the manager rather than a convention. internal/infrastructure/workflowsteps is the one package both composition roots import: it holds the compiled-in provider set and builds the manager. archied registers it in RunStateStore (the validating side) and in openStateStoreAdapter (the daemon's client), archie-agent in productionWorkerDependencies (the executing side), and archie-messaging at its own root, each before the first resolution. The five hardcoded sites now resolve the injected manager instead: the State Store's workflow-definitions validate closure and encodeWorkflowDefinitions, the control-plane client's read and write paths, and CompilePinnedWorkflow's shipped -definition and pinned compiles (one resolution per package). step_vocabulary_agreement_test.go registers a provider through the production Manager and drives all three sides with it; breaking any one of the six resolution points back to the builtin registry fails that side's assertion. |
||
|
|
02720c8e81 |
feat(controlplane): runtime administration over the State Store's gRPC server
The Web UI and messaging channels now read and write Archie's runtime settings through one versioned, audited control-plane API hosted on the State Store's existing gRPC server: catalog, query, command, watch. Each feature owns its resource, validation and commands; storage only writes versions and audit records. Workflows move with it. Definitions become portable YAML compiled through a closed, typed step registry and stored as a control-plane resource, so the yaegi interpretation paths (internal/gate, wfeval, wfextract, gateeval, gateextract, skillbuild) are deleted rather than carried. The UI process reads definitions as data: the JSON projection lives in internal/domain/workflow/task so a dashboard decodes the stored collection without linking the workflow engine, which cmd/archie-ui's deletion gate forbids. Shipped definitions ride the catalog descriptor rather than a bespoke REST route, keeping the API to its four operations. archie-messaging gains -state-store-target/-state-store-token and reads channel settings from the database, failing closed when the State Store is unreachable: the database is authoritative for those settings. |
||
|
|
4750ac097b |
feat(workflow): post line-anchored review comments with suggested fixes (#891)
* feat(workflow): post line-anchored review comments with suggested fixes Bead archie-core-q9au (docs/prds/inline-review.md). A review's findings reached the operator only as a flat list in the PR body, so the author had to map every finding back to its line by hand and apply mechanical fixes by hand. Now the line-anchored findings are also posted as inline review comments, carrying a one-click suggestion when the reviewer could state the fix. Domain (internal/domain/workflow): - ReviewFinding gains Suggestion; Validate rejects one on anything but a confirmed finding, so the agent's record_finding call is refused with something it can act on. Validate splits into validateAnchor / validateClassification / validateDisposition, which also brings it back under the cyclop limit the gate enforces. - StagePostReviewComments runs after StageOpenPR in the implement workflow. It posts only line-anchored findings (Line > 0); a whole-file finding has no line to attach to and stays in the body list, which remains the complete record. It is best-effort: the PR is open and the body already lists every finding, so a failed comment is logged, never a park -- the same reasoning as OpenPR's existing best-effort LinkBranch call. - The suggestion fence is gated on Verdict == confirmed where the fence is written, because it is a one-click apply: a wrong suggestion breaks the code and the author has to notice. Forge (internal/forge): ReviewCommentWriter + GitHub/Gitea implementations, type-asserted like PullRequestReader so a forge that cannot is skipped rather than faked (NoopForge is pinned as not satisfying it). GitHub loops one CreateComment per comment and joins the failures, so one refused anchor cannot drop the rest; Gitea carries them on a single COMMENT-state review. Two deliberate departures from the PRD's sketch, both recorded in it: - The interface is batch (CreateReviewComments), not one call per comment: Gitea's only inline shape is one submitted review, and the set then crosses NATS once instead of once per finding. - The head SHA is resolved inside each forge implementation rather than by the caller. The stage runs in archie-agent, which holds no forge credentials and has no PullRequestReader; the daemon-side forgerpc handler is the only place that does. Transport (internal/forgerpc): workflow.Forger gains the method, so the agent reaches the daemon over the new create_review_comments subject; the server type-asserts the writer and answers an incapable forge with an error the stage logs. This is the one proxied method whose absence on an image skew does not park a task. The reviewer prompt and record_finding schema ask for a suggestion on a confirmed, mechanically fixable defect only. Regenerating the wfextract symbol table -- the documented way a new exported workflow symbol reaches interpreted stage plugins -- also picked up symbols that were missing from the committed table (the Status* and Source* constants, ReviewCheck, Store, Task, and the _Store wrapper). Nil-guards were re-applied to every wrapper afterwards, as that file's header requires. Evidence: task check exit 0 (incl. 229 UI tests). Three mutations of the new stage logic each fail a test that names the behaviour -- dropping the confirmed-only gate, dropping the Line > 0 filter, and letting a posting failure escape -- so the tests discriminate rather than passing vacuously. * fix(build): drop the Dockerfile pin assertion The guard fetched golangci-lint's go.mod at image build time to fail the build when GOLANGCI_LINT_VERSION and GOFUMPT_VERSION disagreed. It only ever tripped on self-inflicted pin drift, but it added a network fetch and a new failure mode to the release build, which is not where that belongs. The pins stay correct (golangci-lint 2.13.2, whose vendored gofumpt v0.11.0 equals GOFUMPT_VERSION) and the comment above them still states the invariant, so the reasoning survives without the build-time machinery. The pin change itself reaches the agent image through the normal release path: CI rebuilds and pushes archie-agent on an archie/vX.Y.Z release, and a gateway-only release deliberately does not. * feat(workflow): add the remediate workflow for PR review remediation Implements step 5 of docs/prds/pr-review-remediation.md: a remediate workflow that resumes the task's existing worktree onto its PR branch (worktree.Manager.Resume, not Prepare's refresh), runs a builder agent against the review's actionable comments, commits and pushes to the same branch, and replies via the forge's ReplyToReview. Task.RetryCount doubles as the remediation round counter, capped by the repo's max_retries; the round-cap stage parks the task and posts one comment when it's hit. Task gains a ReviewPayload field (JSON-encoded review unit) with full store/proto/staterpc plumbing. workflow.Forger gains Comment and ReplyToReview, proxied to archie-agent via a new forgerpc RPC. Resume and its Dir lookup are added as an unexported optional Trees capability rather than widening the exported interface, which is already at golangci-lint's interfacebloat cap and is projected into wfextract's Yaegi symbol table for repository-authored stage plugins. The daemon-side reaction consumer that dispatches into this workflow (subscribing ARCHIE_REACTIONS, grouping comments by review, enforcing per-task serialization) is tracked separately in archie-core-8li9.4. * feat(messaging): extract standalone messaging service binary and process composition * feat(messaging): adapt channel frontends to consume Gateway ChatContract * feat(messaging): remove in-process channels and Router from daemon bootstrap Telegram, email and webhook now run only in the Messaging Service, which reaches this process through the Gateway's ChatContract. The daemon builds no channel and no gateway.Router, so the wiring that existed only to serve them goes with it: setupGateways, setupEmailGateway, setupWebhookGateway, setupTelegramGateway, buildTelegramRouter, makeTelegramReload and the release-announcement path helper. The seams those paths owned move to where they can still be honoured: - Channel ValidateConfig runs in archiemessaging.compose, which now returns an error, so a configured-but-invalid front-end fails the service rather than being dropped silently. - The inbound rate limiter moves to the Gateway process, which owns the only Router and therefore the only place a per-sender budget applies to every channel's turns. - The daemon's gateway readiness probe becomes a ContractProbe against the Gateway. It cannot observe a channel lifecycle it no longer runs, and a channel-count probe would have reported every daemon degraded. - telegramSetup is now chatSetup (in chat_setup.go), trimmed to the fields a chat turn runner actually reads. Telegram operator features that were wired only in the daemon (/update, release announcements, version reporting, SIGHUP reload) are not yet wired in the Messaging Service: archie-core-8cda.6.6. Channel lifecycle reporting: archie-core-8cda.6.8. The Gateway process builds no worktree manager, so review_pr is unadvertised on every channel: archie-core-8cda.6.7. * feat(messaging): re-wire the Telegram operator surface in the Messaging Service setupTelegramGateway satisfied eight telegram.Gateway seams from the daemon's own process. Each is resolved by where the fact it needs is owned, not by where it used to be built; the decision table is in docs/architecture/migration-decisions.md. Wired here: /update (built from this service's own [chat.telegram] check and install commands), the update-report relay (same sha256(bot_user)[:8] path the daemon wrote, so a pending report is found rather than written afresh), show_tool_calls, and /restart's config reload, which re-resolves the token and allowlist from this service's own file and refuses a reload that resolves to an empty token. /version is sourced from the Gateway: ChatSnapshot gains a Version field carrying the Gateway's rendered build block, read per call and bounded by the dependency timeout. cmd/archie-messaging deliberately takes no version ldflags, so it cannot report a partial upgrade as a matched one. RunningVersions and ReleaseAnnouncements stay nil. Both need a component's real running version, and the only versions this process can obtain are a different binary's stamps -- reporting them would manufacture exactly the false success the update check exists to catch, and the announcer without parseable versions is a no-op that reads as configured. archie-core-8cda.6.9 carries the contract that unblocks them. Dangerous stays nil: no composition has ever supplied a command authority. Enrich is dropped from the update service: componentInstallTypeEnricher read [nats], which this service must not decode, so the check command's own reported install type stands. * fix(workflow): guard inline review comments against head drift Two findings from the review of q9au. A comment's line numbers describe exactly one revision of the branch, but the forge resolved the pull request's head at posting time -- after the review had already read the worktree. A push in that window (a collaborator, a retry) meant the findings were anchored to whatever now occupied those lines, mislabelling instead of failing. The revision the review measured now travels with the comment set: OpenPR records it from the same worktree read it already performs (TaskContext.ReviewedHeadSHA), StagePostReviewComments threads it through Forger to forgerpc, and the forge refuses the whole set once the head it reads has moved off it -- which the best-effort stage logs, the PR body remaining the record. An unmeasured revision posts unverified rather than dropping every finding, so a worktree that cannot report one degrades instead of regressing. Suggestion was documented single-line only, because GitHub applies a suggestion fence up to its first line and a multi-line replacement would silently corrupt the file. Only the verdict was validated, so a multi-line suggestion reached the renderer and posted as that truncated fence. Validate now rejects an embedded newline, which makes record_finding refuse it with feedback the model can act on instead of the reviewer discovering the corruption later. |
||
|
|
74682465ca |
chore(plugin): delete the unwired capability-host health surface (#882)
* chore(plugin): delete the unwired capability-host health surface
plugin.Host.Health, plugin.Host.Manifests, ModuleStatus and their private
safeHealth helper were implemented, exhaustively tested, and called from nowhere
but their own tests. AGENTS.md forbids speculative surfaces, and no settled
design places plugin module status anywhere that exists:
- docs/prds/status-health-surface.md Phase 2 is self-described as "a proposed
resolution, awaiting Sam's sign-off -- not yet implemented", and does not
mention plugin status even in that unapproved list.
- The readiness subsystem enumerates exactly five probes -- state_db, config,
disk, model, gateway (internal/app/archied/readiness.go) -- with no plugin
probe.
- docs/architecture/plugins-and-extensions.md names no operator-visible plugin
status surface.
- The only bead naming it is closed.
plugin.Host itself stays: it is the daemon's live capability host, constructed
and Start/Stop-ed from internal/app/archied. Only the unconsumed observability
methods go.
host_contract_test.go encoded the defect -- it asserted plugin.Host *defines*
"Health" and "Manifests" -- so those two names are removed from its list, which
is now exactly the shipped lifecycle: Register, Start, Stop. The three tests
covering the deleted surface are removed with it.
The Yaegi symbol table (internal/plugin/pluginextract) was regenerated rather
than hand-edited, so ModuleStatus leaves the reflected type set with it.
Reachability evidence: Host.Health and Host.Manifests appear only in their own
tests; plugin.Host is otherwise reached from bootstrap.go and gateway.go for
Register/Start/Stop. Note that "task reachability" does NOT flag these -- the
address-taken constructor is a known blind spot of the whole-program RTA.
* chore(plugin): delete the dead Module.Health chain wired to nothing
Extends the capability-host observability deletion: plugin.Module.Health
was a required interface method with real implementations in
tools/provider/mcp and tools/provider/builtin, aggregated by
tools/provider.Registry.Health via a provider-package-local safeHealth,
and backed by per-module LifecycleState bookkeeping in plugin.Host
(registeredModule.state, setModuleState). None of it had a caller before
this change -- Host.Health was the only thing that would ever have read
a registered module's Health(), and Host.Health was already deleted as
unreachable. registeredModule.state was write-only: nothing ever read it.
Deleted:
- plugin.Module.Health, plugin.Health/HealthStatus (+ consts),
plugin.LifecycleState (+ consts), legacyModule.Health,
registeredModule.state, Host.setModuleState and its call sites.
- tools/provider.Registry.Health, tools/provider.safeHealth.
- tools/provider/mcp.Provider.Health, tools/provider/builtin.Provider.Health.
- Tests exclusively covering the above, and fakes' now-unnecessary Health
methods (plugin/host_test.go, pluginextract/reachability_test.go,
tools/provider/{provider,optional}_test.go, tools/provider/mcp/provider_test.go,
app/agentworker/mcp_providers_test.go).
internal/plugin/pluginextract's generated file was regenerated via
`go generate ./internal/plugin/...` and the yaegi wrapper nil-guards
(dropped by fresh generation, as documented in reachability_test.go)
were reapplied by hand to WManifest/WStart/WStop/WName/WVersion,
following this repo's established post-regen pattern.
|
||
|
|
2c09f42ff4 |
feat(tasks): attribute events, changes and config to the attempt that produced them
The events table had no attempt column, so two attempts of the same task were indistinguishable: a retried task's stage events interleaved into one timeline and nothing could say which run produced a given result. - events gains an `attempt` column through the existing presence-gated migrateTasks loop rather than a version bump, so user_version stays 3 and an existing archie.db is migrated in place. A legacy-database test builds the pre-change schema and proves both that existing rows survive unattributed (attempt = 0) and that a new value round-trips. - Event.attempt = 11 and ReadTaskLogRequest.stage = 9 are field additions to existing messages: no new RPC, no new table, and no change to any canonical error message that crosses the state-store wire. - Every task event the workflow engine emits now carries the attempt. Producers that have no attempt yet (task_queued, binding_dispatch_failure) stay unattributed rather than guessing one. - Changed files are captured as a durable changes_captured event at commit and push time, reusing the existing three-dot diff primitive: paths from FilePatches(), counts from Stats(), base_sha recorded as the merge base actually diffed rather than the base tip, a 200-file cap with totals kept complete. Consumption goes through an unexported optional interface so workflow.Trees -- the yaegi capability surface for interpreted stage plugins -- is not widened. A capture is also taken once the pull request exists, so the capture can link it. - The per-attempt configuration is captured as a durable config_captured event carrying the non-secret config.TaskConfig, instead of a new table plus two new state-store RPCs: events are already this system's per-attempt provenance stream, so a table would have given one feature two provenance mechanisms. Redaction is proven against canary secrets. Measured payload: 539 bytes typical, 2333 bytes at 20 models. - Captures carry a schema marker and are refused unless their key set matches the declared payload shape, so drift surfaces as "recorded but could not be read" rather than as a capture of zeroes. - Three read endpoints serve the attempt rail, the changed-file captures and a raw debug view, mapping to UI-owned response types rather than dumping store types. - The task-log read gains a stage filter, bound to the stage logger for that stage's own lines; agent and tool output records no stage. A task-scoped grant may already insert events on its own task, so this provenance is exactly as trustworthy as the rest of that task's event stream. That limit is documented rather than papered over. |
||
|
|
bcf7622a8c |
fix(agentworker): hand the worktree back on every sandbox exit
A container run only restored ownership of the bind-mounted worktree on the CommitAll/Push success path (archie-core#520). The agent runs as root inside the container, so every other way a run can end -- a gate failure that parks the task, a stage error, an early return -- left the worktree owned by UID 0 on the host. The daemon is the process that later cleans and resets that directory as its own non-root user, and it cannot even unlink inside a root-owned directory, so the next attempt failed before it could start no matter what the worktree refresh did. runTask now defers the same reconcileOwnership the commit and push paths already call, so it runs at every exit. The retry recoveries in worktree package cannot fix this from the daemon side: a non-root process cannot chown, and removing the directory needs write access to it. Observed through a supplementary group -- a non-root process may move a file it owns into a group it is a member of -- so the test proves the chown ran without needing root. Residual, filed as archie-core-zcg8: a container that is SIGKILLed never runs this defer. |
||
|
|
0b3dc6c01f |
refactor(forge): delete the unwired RepliesAfter surface
DELETE: IssueForge.RepliesAfter and its Reply type had no production caller -- the waiting_human status exists and is reachable, but automatic detection of a human reply never landed, so the intended poller was never built. Neither implementation paginated either (both read one page of 50), so the method would have been wrong if wired as-is. The manual path is untouched: waiting_human remains reachable and an operator requeues through the existing task-action flow. Removing the interface method also removes seven test fakes that existed only to satisfy it. No behavioural test accompanies this: a pure deletion has no assertion to make fail, and the compiler enforces completeness across every implementation. |
||
|
|
d7b809e61b |
fix(agentworker): wire configured tool-result policy into worker's LoopRunner
TaskConfig now carries the daemon's ToolPolicy (MaxResultChars/SpillDir) via ForTask/ToConfig, and runTask applies it to the constructed LoopRunner. Previously the worker always built the agent loop with a zero-value ToolLimits, so tool results in every task's agent stage were never capped or spilled despite the configured 50k default (agentexec-1, tools-core-11). |
||
|
|
669f98a059 |
fix(daemon): carry kind/label routing bindings to the process that routes
An operator's kind/label -> workflow overrides were decoded, validated and installed into the DAEMON's package-level state via workflow.SetKindWorkflows/SetLabelWorkflows -- but workflow.Route's only production caller is app/agentworker/task_execution.go, in the separate archie-agent process. The bindings never crossed the boundary, so every configured override was silently ignored and routing always fell back to the built-in defaults (bug -> tdd, feature -> feasibility, bootstrap -> bootstrap). Carry the resolved bindings on taskrun.Request, mirroring how MCPServers already crosses this boundary, and apply them in the worker before Route is called. Resolved values are carried rather than file paths, so the worker does not need the routing files to be readable in its sandbox. nil still means "use the built-in defaults", so this is backward compatible for any request built by an older daemon. routeTask is a named seam for that install-then-route step. It mutates package-level routing state, which is sound here because each archie-agent process is task-scoped and ephemeral; if the worker ever served more than one task that assumption would need revisiting. Tests are mutation-verified: removing either the two Set calls in routeTask or the two request fields in runViaAgent reproduces the exact assertion failures the lane observed red-first, so neither test is vacuous. Found by an adversarial package sweep (temp tracker: workflow-engine-1, .local/issue-tracker/). |
||
|
|
bdf73ed371 |
feat(workflow): review calibration -- checked properties, style nits out of contract
Closes archie-core-h019.5. The shipped contract already made only a confirmed
error-level finding block, and kept zero-findings distinct from not-run; what
was missing was the other half of calibration, and that is what a reviewer that
cries wolf needs.
- ReviewReport gains Checked []ReviewCheck {Property, Evidence}: the properties
the reviewer verified clean. Evidence is required, for the same reason a
finding requires a failure scenario -- a check without evidence is an
assertion, not a verification.
- The reviewer records checks through a typed record_checked tool, mirroring
record_finding, so it cannot end its turn having silently skipped the
checklist.
- Style/formatting/naming nits are stated out of contract, in the ReviewFinding
type doc and the reviewer prompt: they are the padding that trains an
operator to skim.
- The PR-body findings section renders the cleared properties, so a
zero-finding review arrives backed by what was checked rather than reading
the same as a review that never looked.
Design: docs/prds/review-calibration.md.
|
||
|
|
ad1e6ea0d3 |
feat(gateway): operator-triggered PR review with selectable reviewer model
Add a channel-neutral review_pr command that runs the adversarial reviewer against an existing pull request and returns structured findings. The Web UI chat and Telegram invoke the same command; the reviewer model is models.reviewer (falling back to models.builder), configurable independently of planner/builder. - forge: PullRequest + PullRequestReader.GetPullRequest (github, gitea) - worktree: Manager.CheckoutPR materialises a PR head for Diff/Snapshot - agentworker: NewOperatorReviewer + ReviewerModel (single model-resolution rule) - gateway: review_pr tool, ChatPRReviewer, wired into TurnRunner for every channel - archied: prReviewer adapter with identity-scoped repo allow-list and per-PR single-flight, so duplicate/concurrent reviews are deterministic Closes archie-core-h019.8. |
||
|
|
05529d48d5 |
refactor(state-store): cut daemon/Gateway over to the remote State Store (archie-core-8cda.4.6) (#774)
* refactor(state-store): delete NATS storerpc; agent store is gRPC-only (archie-core-8cda.4.6) The agent is fully on the gRPC State Store contract, so the legacy NATS storerpc path is dead (docs/prds/state-store-contract.md §12 step 4): - delete internal/storerpc (Server, Client, subjects) and its tests - drop the storerpc registration from registerTaskRPCServers and the now-unused store param - make the agent transport's State Store URL required and Store() return only the gRPC staterpc.Client (no NATS fallback) - rework agentworker/agent transport tests to back the store with a real gRPC State Store server instead of the deleted NATS storerpc server The agent binary now fails fast when STATE_STORE_URL is absent. Agent store calls (workflow.Store) go over gRPC to the archie-state-store service. * refactor(state-store): delete daemon/gateway in-process store ownership (archie-core-8cda.4.6) The daemon and gateway no longer open archie.db in-process: the single SQLite file is owned by the standalone archie-state-store process, and both consumers dial its gRPC State StateStore contract (docs/prds/state-store-contract.md §12 step 7, no dual-store ownership). - openStores no longer opens the production task store or seeds b.st; it only wires secrets/forge and the gateway conversation store - openStateStoreAdapter now requires [services.state].target (the old local default is deleted) and resolves b.stateStore as the remote *staterpc.Client - delete the in-process startStateStoreServer (the multiplexed gRPC server and its per-task bearer-token registry) from the daemon composition - RunGateway now resolves the same remote State Store adapter; its session SQLite stays Gateway-owned and untouched - daemon.ConnectedStateStore carries the configured target plus the resolved bearer token, which containerEnv injects as STATE_STORE_URL / STATE_STORE_TOKEN so archie-agent authenticates to the same remote service - readiness store probe reads b.stateStore PRD-consistent choice (noted): with the in-process server gone, per-task token generation is removed; the daemon/gateway/agent all present the single [services.state].target_token (or STATE_STORE_TOKEN secret) that the standalone store validates. * docs(webui): require [services.state].target in config and deployment templates (archie-core-8cda.4.6) After the in-process store path is deleted, archied and archie-gateway dial the standalone archie-state-store gRPC service via [services.state].target and never open archie.db themselves (docs/prds/state-store-contract.md §12 step 7). - config.example.toml documents [services.state].target as required, with the loopback vs Docker-bridge gateway + target_token rule - deployment templates set [services.state] per topology: bridge gateway + token for container-mode agents, loopback for host-only - deployments/README.md and systemd-user-service.md describe start commands for both topologies and drop the stale "not yet consumed by the daemon" note * style: gofumpt alignment for State Store daemon literal; appease noctx/staticcheck (archie-core-8cda.4.6) * docs(state-store): refresh stale in-process references post-cutover (archie-core-8cda.4.6) |
||
|
|
576e22b554 |
feat(state-store): generalize storerpc as the State Store gRPC contract (archie-core-8cda.4.2) (#770)
* feat(state-store): generalize storerpc as the State Store gRPC contract Implements archie-core-8cda.4.2 against docs/prds/state-store-contract.md rev. 2c: generate protobuf/gRPC client and server adapters for every ratified store contract (proto/state/v1/state.proto, 40 RPCs, one StateStoreService), add internal/infrastructure/staterpc with a bufconn conformance suite exercising both local and remote adapters, serve the service in-process (bridge address + mandatory per-task bearer token when agent containers exist, loopback-only otherwise), and flip archie-agent's workflow.Store to a long-lived gRPC client when the daemon injects STATE_STORE_URL/STATE_STORE_TOKEN (falling back to the legacy NATS storerpc path otherwise -- one agent process is never both). Also drops the *sql.Tx parameter from store.BindingDispatcher.RecordDispatch (a required wire-blocker fix: a transaction cannot cross a gRPC boundary, and production always passed nil). fix(store): drop RecordDispatch's unusable *sql.Tx parameter feat(config): add services.state contract-adapter selector * fix(state-store): apply no-mistakes review decisions - Drop StreamTokenInterceptor (unused; the State Store service has no streaming RPCs and nothing calls it). - Make StateStoreTokens.Validate compare tokens in constant time (crypto/subtle) per the auto-fix hardening.ing finding. - Amend config.example.toml / config.Services.State doc text so it frames [services.state].target as the ratified surface consumed by the upcoming standalone archie-state-store binary (.4.3), not active today. |
||
|
|
5c0efaceaf |
refactor(workflow): relocate Task/Status/Source and define workflow.Store (#754)
* refactor(workflow,store): relocate Task/Status/Source and define workflow.Store Move store.Task, store.Status*/store.Source*, and the 3-method workflow store interface out of internal/store into internal/domain/workflow as workflow.Task/workflow.Status/workflow.Source plus a consumer-owned workflow.Store, superseding store.WorkflowStore. internal/domain/workflow now imports zero internal/store symbols; internal/store persists the domain type instead (infrastructure depends on domain), and its Store satisfies workflow.Store. Updates every production and test caller of the relocated symbols across daemon, webui, gateway, agentworker, agenttransport/nats, storerpc, taskrun, and worktreerpc, and removes TestRecordDispatchViaExplicitTx per docs/prds/state-store-contract.md (rev. 2c) sections 4 and 12 step 2. Two workflow-package integration tests that exercise the real *store.Store move to an external workflow_test package (store_integration_test.go) to avoid a workflow(test)->store->workflow(prod) import cycle now that store depends on workflow. * no-mistakes(document): docs(state-store): mark §4.8 relocation done in state-store-contract |
||
|
|
c3e9e4c43e |
feat(workflow): adversarial self-review stage before PR open
Implements archie-core-h019.3 (reviewer isolation) and h019.4 (wiring),
per docs/prds/adversarial-self-review.md.
- internal/domain/workflow: new Reviewer interface and StageReview,
inserted between StageDiffCap and StageOpenPR in the implement
workflow. Fail-closed: a review that didn't run, or a confirmed
error-level finding, parks the task (store.StatusParked) with
findings in Detail; warn/plausible findings never block. Inert
unless Repo.ReviewEnabled.
- internal/worktree: Manager.Snapshot exports HEAD's tracked files into
a fresh directory with no .git, for the reviewer's isolated
workspace. Trees interface gains Snapshot; hybridTrees delegates to
it.
- internal/app/agentworker: subagentReviewer implements Reviewer using
ai-sdk's agent.Subagent (nested, in-process, non-streaming
generation) -- conversation isolation is structural (Subagent.Run
never carries message history), workspace isolation is the snapshot
above. The reviewer's toolset is read-only (read/grep/find) rooted
at the snapshot, never the worker's own toolset, so it cannot reach
the worktreerpc publication grant. Findings are captured via a typed
record_finding tool appending to a slice the stage owns, read even
when Subagent.Run errors (a truncated run's captured findings
survive).
- internal/config: Repo.ReviewEnabled (default false) and
Models["reviewer"] (falls back to Models["builder"], no schema
change).
Judgment calls: Reviewer.Review never returns a Go error -- every
failure folds into ReviewReport per the PRD's fail-closed contract,
which the stage can act on uniformly. A Subagent.Run error with zero
captured findings resolves to ReviewStatusNotRun (park); a
Subagent.Run error with >=1 captured findings resolves to
ReviewStatusCompleted with those findings, since the PRD's own example
("three defects survive a step-limit truncation") requires them to
still be evaluated for blocking rather than discarded.
|
||
|
|
a825db283c |
fix: harden task and chat execution
Task runs now preserve real model context limits, provider usage, baseline-repair spend, commit identity, durable agent-finish telemetry, and useful timeline details without imposing a token cap. Dashboard forge links are canonical per owning GitHub or Gitea identity. Chat turn claims acquire SQLite writer intent before reading, eliminating the concurrent SQLITE_BUSY snapshot race. |
||
|
|
694a977ffe | fix(agentworker): reconcile task clone ownership after commits | ||
|
|
5d00ec7b0c |
refactor(agentexec): remove legacy execution paths
Autonomous workflows now have exactly one production boundary: a task-scoped archie-agent receives a full-task core-NATS request and runs the worker-local stage loop. The legacy [agent] section remains decode-only for migration; it no longer selects runtime behavior. |
||
|
|
73a83a133b | chore: apply gofumpt formatting | ||
|
|
ed33ab5380 |
feat(agent): stamp archie-agent's build and report it back to the daemon
archied's own build has carried a compiled-in gatewayVersion and an internal/installtype stamp since archie-core-522/b5d08d0, but archie-agent's build had neither -- its Taskfile.yml and Dockerfile build steps passed no -ldflags at all, so even a daemon that wanted to verify what agent build is actually in service had nothing to check against. daemonRunningVersions said as much explicitly: the agent was left out of every verification on purpose, because nothing self-reported a real value. Every archie-agent process is task-scoped and ephemeral (spawned per task, dies after), not a long-lived process the daemon can poll on demand the way it polls Telegram's gateway. So there's no "ask the agent its version" request -- the only channel back is whatever a worker includes in the taskrun.Response it already sends on every completed task. Stamp archie-agent's build the same way archied's is (installtype, plus a new agentworker.Version()), have executeTaskRequest report both in its Response, and have the daemon record the latest observation in a new mutex-guarded AgentStatus, fed from runViaAgent on every successful response only -- never off a request that never reached an agent, or one describing a failed/parked task. daemonRunningVersions now includes the agent component once (and only once) something has actually been observed, so Report.Verify gets a real value to check future update claims against instead of omitting the component entirely. Part 1/3 of archie-core-1786751948782-3-837a8fd0 point 4. Closes archie-core-1786751948782-3-837a8fd0.1. |
||
|
|
537651fb4c |
Merge pull request #562 from samcharles93/fix-agentworker-json-import
fix(agentworker): move JSONL decode into storage, unblocking CI |
||
|
|
ab5a60c1b8 |
fix(agentworker): move JSONL decode into storage, unblocking CI
TestApplicationOwnsNoTransportWireMechanics was failing on main since
|
||
|
|
dbb6550ab3 |
fix(worktree): bind publication to dispatch grants
Worktree RPC used to accept repository coordinates from the sandbox. It now accepts only an opaque per-dispatch grant, and the daemon derives the repository path and branch from grant-owned state. Repository worktree paths also move from an ambiguous flattened key to encoded owner/repo components. Origin-matched legacy clones migrate safely; mismatched legacy paths are preserved. |
||
|
|
99738845b6 |
fix(container): persist npm cache for per-task MCP servers
Ephemeral task containers re-resolve and re-download npx-launched MCP server packages from the npm registry on every task, since each task gets a fresh container with no persistent cache -- a worse amplifier of a registry outage than the daemon-restart case d4bb914b already fixed. Extract the daemon's npm-cache env logic into mcp.NpmCacheEnv, shared by both the daemon's own long-lived npx MCP servers and archie-agent's per-task ones. Mount a new archie-cache-mcp-npm Docker volume at /data/cache/mcp-npm on every task container regardless of ecosystem (MCP servers are daemon-wide config, not tied to the task repo's language), and point npx-launched per-task stdio transports at it. Closes archie-core-qk8m. |
||
|
|
d7801bb7d3 | feat(storage): persist agent project state | ||
|
|
5da1d5d006 |
fix(agentworker,daemon): reconcile worktree ownership before push
Agent containers run as root (no USER in the Dockerfile, no userns-remap), so a commit the agent writes to the bind-mounted worktree lands owned by UID 0 on the host. The daemon then reads those same loose objects to push, running as its own non-root host user, and fails with "openat objects/../..: permission denied". containerEnv() now passes the daemon's own UID/GID as WORKTREE_UID/WORKTREE_GID. hybridTrees.Push -- the single point where a commit's objects are handed back for the actual push -- chowns the worktree to that owner first, via os.Lchown so a legitimately-tracked-but-dangling symlink doesn't abort a push that would otherwise have succeeded. Closes archie-core-1786749700606-6-bf6aa32d. |
||
|
|
16ec8393fe | style: apply gofumpt formatting | ||
|
|
20bff1a367 |
feat(agentexec,workflow): surface tool calls on the task timeline
archie-core-467 wired tool-call visibility into the interactive chat gateway, but a workflow task's own transcript never showed what its tools actually did -- only stage/agent-finish events existed, so there was no way to confirm an agent's summary matched what it really ran. Adds a ToolCallReporter, invoked once per completed call from toolExecute (the single choke point every tool call passes through before reaching a model), reported via a new ToolSetOptions.OnToolCall hook. Runner.Run grows a report parameter to carry it in: InProcessRunner wires it through, since it executes in the same process as its tools; SubprocessRunner and NATSRunner accept and ignore it, since their tool calls happen in a different process with nothing here to observe them. The workflow engine turns each report into a tool_call event on the task's own Bus (agent.go, implement.go), riding the same NATS event-forwarding archie-core-518 already built to get container events onto the daemon's timeline. Scoped to the central tool registry (read/write/edit/grep/shell/...) built via BuildToolSet; capture tools (structured finish output) and run_go_script build their ToolSet entries directly and are not yet covered. |
||
|
|
a237a35532 |
fix(agentworker): back off before refetching a nak'd stage message
On a stage-handle failure, the main loop logged the error, called Nak (which requests redelivery), and looped straight back into FetchStage with no delay -- unlike the fetch-error branch three lines above, which already backs off for a second. A deterministic failure (a broken gate, a malformed request) therefore refetched and failed again immediately, flooding the log with one "handle failed" line per loop iteration instead of per genuine retry -- hundreds of identical lines within the same second in production. Adds the same time.Second backoff after Nak. |
||
|
|
d9a9b17153 |
fix(agentworker): forward workflow events across the NATS worker boundary
A workflow run through the container/NATS path never wired workflow.TaskContext.Bus, so every tc.Emit call (stage_start, stage_finish, outcome, parked) was a silent no-op. archied's own in-process path sets Bus and always worked; anything dispatched to an archie-agent worker did not, which meant the dashboard's per-task timeline showed "No events yet" for every task regardless of status or stage -- confirmed live across a Working task and two differently parked tasks. Bridges the gap the same way system logs already cross this boundary: ForwardTaskEvents relays a per-task in-process *events.Bus to a new NATS subject (SubjectForEvents), and a new daemon-side subscriber (subscribeAgentEvents) republishes decoded events on the daemon's own Bus, landing them in the same persist-and-broadcast pipeline an in-process workflow already used. Closes #518 |
||
|
|
45345cdfb6 |
refactor(agent): move worker runtime into application layer
cmd/archie-agent used to own worker construction, NATS lifecycle, task execution, MCP setup, and workflow wiring. It now owns process input only and invokes internal/app/agentworker, while transport-specific NATS, boot-brief, and git-safety mechanics live under infrastructure. |