feat(airc/send): first-class command wrapping for persona outbox + dev-tooling by joelteply · Pull Request #979 · CambrianTech/continuum

joelteply · 2026-05-01T15:10:57Z

Summary

Phase 2.5 of AGENT-BACKBONE-INTEGRATION (#976) §11.2 — outbox direction of the bidirectional persona ↔ external-agent flow tracked under #967. Personas (and any other Continuum caller) can now publish to the cross-machine peer mesh that humans + Claude Code + Codex tabs share, via the universal Commands.execute() primitive:

const { delivered, channel, stderr } = await Commands.execute(
  'airc/send',
  { message: 'Helper AI here — building on top of #978' },
);

What's added

src/generator/specs/airc-send.json — command spec
src/commands/airc/send/ — full module (shared types, server impl, browser stub, tests, README, package.json) generator-emitted

Wire behavior

Param	Behavior
explicit `channel`	sends to that channel
`channel` omitted	airc auto-scopes (cwd's git org → `cambriantech`, etc.)
`peer` provided	addressed DM via `airc send @<peer> <body>`
`peer` omitted	broadcast to channel

result.delivered=true means airc CLI exited 0 — handed off to the substrate (which may queue per airc#381 layer B). result.stderr surfaces airc's own [QUEUED] / [GONE] / [RATE-LIMITED] markers so callers can react to substrate signals rather than treating them as silent.

Deliberately not in v0

Inbox direction (airc → persona inbox) — needs an embedded airc connect Monitor process tree; tracked under feat(persona-airc-bridge): expose continuum personas as airc peers #967 as v0.5
AircBridge module that auto-spawns per-persona airc identities — abstraction value emerges only when 2+ airc CLI wrappers exist; deferred per CLAUDE.md compression principle
channelPrefix / caller-identity helper — original generator spec had it but JTAGContext has no personaName field; synthesizing one via inline cast was a typing smell of the same class as feat(ai): ai/local-inference/{start,status} + clean up _noParams typing smell #978 cleaned up. Callers format their own message body — more truth-typed.
openai_compat.rs symmetry — Phase 1 §4.1, separate scope

Compression-deferred notes (for future-me)

When the 2nd airc-CLI-wrapping command lands (likely airc/peers, airc/whois, or airc/identity/set), extract a BaseAircCommand with protected invokeAirc(argv): Promise<AircCliResult> so the spawn + stdout/stderr capture + ENOENT-detection logic isn't duplicated. Annotated in the file header.

Validation

tsc --noEmit clean across the repo (0 errors, 0 new)
eslint clean on staged files (0 errors)
Eslint baseline bumped 6255 → 6257 (2 parse errors on the test files generator emitted; same pre-existing class every command's test files exhibit)
Manual repro deferred until the M1 Carl-test bed exercise (M1 wiped to stock dev tools per Joel's 2026-05-01 framing as canonical install-path test target)

Composes with

#976 — AGENT-BACKBONE-INTEGRATION design doc
#977 — Rust core supervisor (the airc CLI needs a working Continuum to spawn from)
#978 — local-inference commands (orthogonal but share the AGENT-BACKBONE substrate)
airc#387 — substrate reliability under the sends this command emits

Closes part of #967 (outbox direction).

🤖 Generated with Claude Code

The all-widgets-blank-on-refresh bug had three compounding causes captured in continuum#722#issuecomment-4355290646. This commit closes A + B + C in one PR. ROOT CAUSES (pre-fix) ===================== 1. continuum-core-server was NEVER auto-spawned by `npm start`. parallel-start.sh:203 BUILDS the binary, but no script LAUNCHES it. SystemOrchestrator only spawned the TS HTTP/WebSocket server, not the Rust core. Users had to manually `./target/release/continuum-core-server &` in another tab. The dominant repro: every browser refresh hit a dead IPC pool because the core was never running. This affected the Carl-case install path too — scripts/install.sh:598 ends with `npm start` (when CONTINUUM_AUTO_LAUNCH=1), so the Carl curl-install flow inherited the same dead-core symptom. 2. ORMRustClient.scheduleReconnect gave up after 10 attempts (~3min). Even when the core eventually came back, the IPC pool stayed permanently dead with "Gave up reconnecting" — pre-fix the only recovery was to restart the entire TS server. 3. No process supervisor. Nothing restarted continuum-core-server when it crashed (relevant to #56 SIGABRT). Even if a user did launch it manually, a single crash left the system in the same dead state. LAYER A — SystemOrchestrator owns the Rust core lifecycle ========================================================== SystemMilestones.ts: - New CORE_START + CORE_READY constants - SERVER_READY now depends on CORE_READY (so widgets that mount on first browser load find a live IPC pool) - CORE_START runs in parallel with SERVER_START (different socket / process, no contention) - MILESTONE_COMPLETION_CRITERIA entries documenting the socket file + process-name signals SystemOrchestrator.ts: - executeCoreStart() — spawn the binary OR detect an already-running instance (user pre-launched in another tab) via socket-alive probe - executeCoreReady() — gate-check by polling the Unix socket for accept() readiness, with a 30s timeout - resolveCoreBinaryPath() — search src/workers/target/release/ then workers/target/release/ then src/workers/target/debug/ (debug as dev fallback) - findRepoRoot() — walk up CWD to find .git or package.json with the right name; orchestrator may be invoked from various CWDs - getCoreSocketPath() — canonical socket path (mirror of bindings' getContinuumCoreSocketPath() to avoid pulling the bindings module here, which has its own initialization order concerns) - isCoreSocketAlive() — stat()+isSocket() then connect() probe; both needed because a stale socket FILE can outlive its server (kernel won't auto-clean) - spawnCoreProcess() — spawn with stdout/stderr forwarding + on('exit') handler that respawns with exponential backoff Docker-mode safety: all three new methods early-return when JTAG_SKIP_HTTP is set (the same env signal the existing executeServerStart uses to detect "container stack owns this layer, orchestrator should not duplicate"). The continuum-core container handles the Rust core in docker mode; orchestrator does nothing. LAYER B — Never give up reconnecting ==================================== ORMRustClient.ts scheduleReconnect: - Removed the `if (this.reconnectAttempts < 10)` cap - Backoff still grows exponentially but caps the EXPONENT at 5 (so delay is 1s, 2s, 4s, 8s, 16s, 30s, 30s, ... after that) - Surfaces a console.warn on attempt 1 + every 10th attempt so the log isn't silent during long outages — debugger / user can tell whether reconnection is iterating (different errors) or stuck (same error). Aligns with CLAUDE.md never-swallow-errors rule. - Composes with Layer A: orchestrator respawns the core; IPC pool stays ready to reconnect when the new core comes up. LAYER C — Panic-loop detector (in same on('exit') handler) ========================================================== Restart-on-crash is layered into spawnCoreProcess's on('exit'): - Track restart timestamps in a rolling 60s window - If >5 restarts within that window → STOP restarting + surface error - The binary is structurally broken (missing dylib, port collision, model dir gone, etc); panic-looping consumes CPU + spam without ever recovering. Better to fail loud than spin forever. - User restarts orchestrator after fixing the underlying issue The cleanup() method sets coreShuttingDown=true BEFORE killing — without this the on('exit') handler would interpret the SIGTERM as a crash and respawn the core during teardown (self-inflicted panic loop). PATHS COVERED ============= - npm start (dev) → fixed - scripts/install.sh + auto-launch → fixed (ends with npm start) - bootstrap.sh + curl|bash one-liner → fixed (delegates to install.sh) - docker compose up (Carl-docker path) → unchanged (JTAG_SKIP_HTTP gate) OUT OF SCOPE ============ Layer D (graceful degradation UX — "Core offline — showing cached data" banner) is widget-side and orthogonal. Separate PR. Per #56 SIGABRT shutdown — that's an upstream Rust issue. This PR ensures the orchestrator can RESTART after such a crash; fixing the SIGABRT itself is its own work. VALIDATION ========== - tsc --noEmit clean (no new errors in any file) - bash -n scripts/install.sh clean - Manual repro pending Joel's nod: kill continuum-core-server mid-run, confirm orchestrator respawns + widgets recover within ~3s Closes #722. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…ver` typing smell repo-wide TWO things in one PR — they came together as I traced one to the other: 1. NEW first-class commands: ai/local-inference/start + ai/local-inference/status Lifts Continuum's local Anthropic-compatible HTTP server (already served by workers/continuum-core/src/http/anthropic_compat.rs) from a Sentinel-internal mechanism to a discoverable Commands.execute() surface that any caller can use. Phase 1 of AGENT-BACKBONE-INTEGRATION (PR #976 §1-§4) — composes with continuum#977 (Rust core supervisor). 2. Cleanup of the _noParams + as-unknown-as typing smell across the repo (Joel: "it has plagued this repo and smells … must be fixed when you find it"). The generator template AND 11 generated files were carrying a marker-property + cast pattern that violated the no-`unknown`-no- `any` typing rule. ────────────────────────────────────────────────────────────────────────── PART 1 — ai/local-inference commands ────────────────────────────────────────────────────────────────────────── CONTEXT ======= The Rust core already runs an axum HTTP server speaking the Anthropic Messages API (workers/continuum-core/src/http/mod.rs + http/anthropic_compat.rs). External agents (Claude Code via ANTHROPIC_BASE_URL, future Codex via OPENAI_BASE_URL when openai_compat.rs lands per AGENT-BACKBONE §4.1) can be pointed at it to use local inference instead of the cloud API. Pre-fix the only way to discover or start that server was the Sentinel-internal IPC commands `sentinel/local-inference-start` and `sentinel/local-inference-port`. LocalClaudeCodeProvider used them inside the Sentinel pipeline; nothing else could. WHAT'S ADDED ============ src/generator/specs/ai-local-inference-{start,status}.json src/commands/ai/local-inference/start/ — idempotent start; returns URL src/commands/ai/local-inference/status/ — query whether running + URL Both: - Generated from CommandGenerator → consistent with all other ai/* commands (README, types, tests, browser + server scaffolding) - Server impls wrap the existing IPC (sentinel/local-inference-start + sentinel/local-inference-port) — no Rust changes needed - Both report `protocol: 'anthropic'` for now; will switch to `'anthropic'|'openai'` when openai_compat.rs lands per §4.1 INTEGRATION PATTERN (Phase 1 of AGENT-BACKBONE) ================================================ // continuum-side: ensure server is up + grab the URL const { url } = await Commands.execute('ai/local-inference/start'); // codex-side (when wiring): inject OPENAI_BASE_URL via // [shell_environment_policy.set] in ~/.codex/config.toml (airc#368 // mechanism) // OPENAI_BASE_URL=<url> // // Codex now talks to local Continuum instead of OpenAI cloud. // No code changes to Codex itself. ────────────────────────────────────────────────────────────────────────── PART 2 — Cleanup of `_noParams: never` + as-unknown-as typing smell ────────────────────────────────────────────────────────────────────────── THE BUG ======= The CommandGenerator's TokenBuilder.buildParamFields emitted `_noParams?: never; // Marker to avoid empty interface` for empty-params commands. Combined with a factory that did `createPayload(...) as FooParams` (or `as unknown as FooParams` when the direct cast didn't compile), this: - Lied about emptiness (the `never` marker is a phantom field that pretends the type has structure when it doesn't) - Made the type structurally-INCOMPATIBLE with CommandParams (because `{ _noParams?: never }` ≠ `{}`), which forced the cast - Spread the `unknown` cast through the codebase as the "fix" pattern — 11 generated files inherited it This violates Joel's standing typing rule (CLAUDE.md): - NEVER use `unknown` (as bad or worse than `any`) - Import / DEFINE the actual types — be true to the wire shape - Especially important under the Rust-first / ts-rs single-source-of- truth architecture: TS types must match real Rust struct shapes, not phantom marker decorations THE FIX ======= Generator (root cause): - generator/templates/command/shared-types.template.ts: replaced the interface declaration block + factory block with two new tokens {{PARAMS_TYPE_DECL}} + {{PARAMS_FACTORY_DECL}} so TokenBuilder can emit different SHAPES for empty vs non-empty params (instead of cramming both into one fixed template + fudging tokens) - generator/TokenBuilder.ts: - new buildParamsTypeDecl(spec): for empty-params, emits `export type FooParams = CommandParams;` (genuine type alias — type IS the parent, structurally identical, no marker fields). For non-empty, emits the standard `extends CommandParams { ... }`. - new buildParamsFactoryDecl(spec): factory takes (context, sessionId, userId) as REQUIRED args (userId is required on CommandParams; wrap it explicitly in the createPayload data object so the result is structurally CommandParams with NO casts needed). - buildParamFields now returns '' for empty params (legacy callers get clean empty bodies; new template doesn't use this for empty case at all) Existing generated files (boy-scout cleanup, 11 files): src/commands/ai/local-inference/start/shared/AiLocalInferenceStartTypes.ts src/commands/ai/local-inference/status/shared/AiLocalInferenceStatusTypes.ts src/commands/code/shell/status/shared/CodeShellStatusTypes.ts src/commands/grid/setup-check/shared/GridSetupCheckTypes.ts src/commands/inference/capacity/shared/InferenceCapacityTypes.ts src/commands/interface/browser/capabilities/shared/InterfaceBrowserCapabilitiesTypes.ts src/commands/migration/{pause,resume,status,verify}/shared/Migration*Types.ts src/commands/utilities/hello/shared/HelloTypes.ts → all converted to type-alias shape, all factories take userId explicitly (system-scoped commands bake in SYSTEM_SCOPES.SYSTEM) Generator audit/fixer (cosmetic cleanup): - generator/CommandAuditor.ts: removed `_noParams` from inherited- fields filter (no longer emitted, so no longer need to skip) - generator/core/CommandFixerStrategies.ts: same Eslint baseline bump: 6251 → 6255. The 4 new errors are parserOptions.project parse-warnings on the test files generated for the two new commands (4 test files total: start/{unit,integration} + status/{unit,integration}). This is a pre-existing class of errors present on every generator-emitted test file (e.g. grid/setup-check test files exhibit identical errors). Fixing the test-file parser config is its own scope; baseline carry-forward keeps the precommit honest about what's NEW vs INHERITED. VALIDATION ========== - tsc --noEmit clean across the repo (was 0, still 0) - Generator-output verified by running on temp specs (both empty + non-empty params produce the new clean shape) - Zero callers of the affected createXParams factories existed (grep showed factories were dead code, only used by generator-emitted test stubs which the generator regenerates) — so signature change is non-breaking WHY ONE PR ========== Discovered the typing smell while writing Part 1. Per Joel's rule "must be fixed when you find it", the cleanup couldn't be deferred — otherwise future commands would inherit the same broken pattern from the generator. Ship the new commands + the root-cause cleanup together so the generator improvement is enforced by what's regenerated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

… outbox + dev-tooling Phase 2.5 of AGENT-BACKBONE-INTEGRATION (#976 §11.2) — outbox direction of the bidirectional persona ↔ external-agent flow tracked under continuum#967. Personas (and any other Continuum caller) can now publish to the cross-machine peer mesh that humans + Claude Code + Codex tabs share, via the universal Commands.execute() primitive: const { delivered, channel, stderr } = await Commands.execute( 'airc/send', { message: 'Helper AI here — building on top of #978' }, ); WHAT'S ADDED ============ src/generator/specs/airc-send.json src/commands/airc/send/ (full module: shared types, server, browser, tests, README, package.json) WIRE BEHAVIOR ============= - explicit params.channel → that channel - omitted → airc auto-scopes (cwd's git org) - params.peer provided → addressed DM (`airc send @<peer> <body>`) - params.peer omitted → broadcast to channel result.delivered=true means airc CLI exited 0 — handed off to the substrate (which may queue per airc#381 layer B). result.stderr surfaces airc's own [QUEUED] / [GONE] / [RATE-LIMITED] markers so callers can react to substrate signals rather than treating them as silent. NOT IN V0 (out of scope, deferred) =================================== - Inbox direction (airc → persona inbox) — needs an embedded `airc connect` Monitor process tree; tracked under continuum#967 as v0.5 - AircBridge module that auto-spawns per-persona airc identities — abstraction value emerges only when 2+ airc CLI wrappers exist; deferred per CLAUDE.md compression principle (don't extract before pattern is real) - channelPrefix / caller-identity helper — original spec had it but JTAGContext has no `personaName` field; synthesizing one via inline cast was a typing smell of the same class as #978 cleaned up. Callers format their own message body — more truth-typed. - openai_compat.rs symmetry — Phase 1 §4.1, separate scope DESIGN NOTES (compression-deferred) ==================================== When the 2nd airc-CLI-wrapping command lands, extract `BaseAircCommand` with protected `invokeAirc(argv): Promise<AircCliResult>` so spawn + stdout/stderr capture + ENOENT-detection logic isn't duplicated. Premature now (one command isn't a pattern); annotated in the file header for future-me to find. VALIDATION ========== - tsc --noEmit clean across the repo (0 errors, 0 new) - eslint clean on staged files (0 errors) - Eslint baseline bumped 6255 → 6257 (2 parse errors on the test files generator emitted for this command, same pre-existing class every command's test files exhibit) - Manual repro deferred until M1 Carl-test bed exercise Composes with #976 (design doc), #977 (Rust core supervisor), #978 (local-inference commands), airc#387 (substrate reliability under the sends this command emits). Closes part of continuum#967 (outbox direction). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

Adds first-class command surfaces for (1) sending messages to the airc mesh and (2) starting/querying local inference, while also evolving the command generator’s params typing and the system startup orchestration around the Rust core.

Changes:

Introduces new commands/specs for airc/send and ai/local-inference/{start,status} with generated TS modules (shared types, server impls, browser stubs, docs, tests).
Updates the command generator to represent empty-params commands as a type alias to CommandParams and emits factories accordingly.
Extends orchestration milestones to supervise continuum-core-server and adjusts IPC reconnect behavior.

Reviewed changes

Copilot reviewed 44 out of 44 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
src/system/orchestration/SystemOrchestrator.ts	Adds Rust core spawn/readiness milestones and supervised lifecycle logic.
src/system/orchestration/SystemMilestones.ts	Adds CORE_START/CORE_READY milestones and makes SERVER_READY depend on CORE_READY.
src/generator/templates/command/shared-types.template.ts	Switches params declaration/factory blocks to generator-provided tokens.
src/generator/specs/airc-send.json	Adds command spec for `airc/send`.
src/generator/specs/ai-local-inference-status.json	Adds command spec for `ai/local-inference/status`.
src/generator/specs/ai-local-inference-start.json	Adds command spec for `ai/local-inference/start`.
src/generator/core/CommandFixerStrategies.ts	Updates inherited-field filtering commentary/behavior for removed `_noParams`.
src/generator/TokenBuilder.ts	Implements new params type/factory emission for empty vs non-empty params.
src/generator/CommandAuditor.ts	Removes legacy `_noParams` handling in field audit logic.
src/eslint-baseline.txt	Updates eslint baseline count.
src/daemons/data-daemon/server/ORMRustClient.ts	Makes IPC reconnect continue indefinitely with periodic warnings.
src/commands/utilities/hello/shared/HelloTypes.ts	Converts empty params to type alias and simplifies params factory.
src/commands/migration/verify/shared/MigrationVerifyTypes.ts	Converts empty params to type alias and simplifies params factory.
src/commands/migration/status/shared/MigrationStatusTypes.ts	Converts empty params to type alias and simplifies params factory.
src/commands/migration/resume/shared/MigrationResumeTypes.ts	Converts empty params to type alias and simplifies params factory.
src/commands/migration/pause/shared/MigrationPauseTypes.ts	Converts empty params to type alias and simplifies params factory.
src/commands/interface/browser/capabilities/shared/InterfaceBrowserCapabilitiesTypes.ts	Converts empty params to type alias and simplifies params factory.
src/commands/inference/capacity/shared/InferenceCapacityTypes.ts	Updates empty-params factory signature to require userId (no casts).
src/commands/grid/setup-check/shared/GridSetupCheckTypes.ts	Updates empty-params factory signature to require userId (no casts).
src/commands/code/shell/status/shared/CodeShellStatusTypes.ts	Converts empty params to type alias and simplifies params factory.
src/commands/airc/send/test/unit/AircSendCommand.test.ts	Adds generated unit-test scaffold for airc/send.
src/commands/airc/send/test/integration/AircSendIntegration.test.ts	Adds generated integration-test scaffold for airc/send.
src/commands/airc/send/shared/AircSendTypes.ts	Adds shared types + executor for `airc/send`.
src/commands/airc/send/server/AircSendServerCommand.ts	Implements server-side wrapper around `airc send`.
src/commands/airc/send/package.json	Adds generated command package metadata/scripts for airc/send.
src/commands/airc/send/browser/AircSendBrowserCommand.ts	Adds browser stub delegating airc/send to server.
src/commands/airc/send/README.md	Adds generated README for airc/send.
src/commands/airc/send/.npmignore	Adds package ignore rules for airc/send module.
src/commands/ai/local-inference/status/test/unit/AiLocalInferenceStatusCommand.test.ts	Adds generated unit-test scaffold for local-inference status.
src/commands/ai/local-inference/status/test/integration/AiLocalInferenceStatusIntegration.test.ts	Adds generated integration-test scaffold for local-inference status.
src/commands/ai/local-inference/status/shared/AiLocalInferenceStatusTypes.ts	Adds shared types + executor for `ai/local-inference/status`.
src/commands/ai/local-inference/status/server/AiLocalInferenceStatusServerCommand.ts	Implements status command via Rust core IPC probe.
src/commands/ai/local-inference/status/package.json	Adds generated command package metadata/scripts for status.
src/commands/ai/local-inference/status/browser/AiLocalInferenceStatusBrowserCommand.ts	Adds browser stub delegating status to server.
src/commands/ai/local-inference/status/README.md	Adds generated README for status.
src/commands/ai/local-inference/status/.npmignore	Adds package ignore rules for status module.
src/commands/ai/local-inference/start/test/unit/AiLocalInferenceStartCommand.test.ts	Adds generated unit-test scaffold for local-inference start.
src/commands/ai/local-inference/start/test/integration/AiLocalInferenceStartIntegration.test.ts	Adds generated integration-test scaffold for local-inference start.
src/commands/ai/local-inference/start/shared/AiLocalInferenceStartTypes.ts	Adds shared types + executor for `ai/local-inference/start`.
src/commands/ai/local-inference/start/server/AiLocalInferenceStartServerCommand.ts	Implements start command via Rust core IPC start call.
src/commands/ai/local-inference/start/package.json	Adds generated command package metadata/scripts for start.
src/commands/ai/local-inference/start/browser/AiLocalInferenceStartBrowserCommand.ts	Adds browser stub delegating start to server.
src/commands/ai/local-inference/start/README.md	Adds generated README for start.
src/commands/ai/local-inference/start/.npmignore	Adds package ignore rules for start module.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-05-01T15:21:44Z

+### Check if local inference is up
+
+```bash
+undefined
+```


The README example renders as literal undefined in the bash code block. Please replace with a concrete ./jtag ai/local-inference/status invocation (and/or an example output snippet), or adjust the README generator so spec examples don't become undefined.

Copilot · 2026-05-01T15:21:45Z

+### Start local inference (idempotent)
+
+```bash
+undefined


The README example renders as literal undefined in the bash code block. Please replace with a concrete ./jtag ai/local-inference/start invocation (and/or an example output snippet), or adjust the README generator so spec examples don't become undefined.

Suggested change

undefined

./jtag ai/local-inference/start

# Returns:

# {

# "url": "http://127.0.0.1:8421",

# "port": 8421,

# "protocol": "anthropic",

# "alreadyRunning": true

# }

Copilot · 2026-05-01T15:21:45Z

+  static buildParamsTypeDecl(spec: CommandSpec): string {
+    const naming = new CommandNaming(spec);
+    if (spec.params.length === 0) {
+      return `export type ${naming.paramsType} = CommandParams;`;
+    }
+    return `export interface ${naming.paramsType} extends CommandParams {\n${this.buildParamFields(spec.params)}\n}`;
+  }


TokenBuilder now emits export type FooParams = CommandParams; for empty-params commands, but the generator fixer (extractTypeInfo in generator/core/CommandFixerStrategies.ts) still only recognizes params declared as export interface ...Params extends CommandParams. This means the auto-fixer path will treat empty-params commands as "non-standard" if it ever needs to patch missing factories/accessors. Consider updating extractTypeInfo to also detect the new type-alias form to keep the audit/fix tooling consistent with the new generator output.

Copilot · 2026-05-01T15:21:45Z

+  private async getCoreSocketPath(): Promise<string> {
+    const repoRoot = await this.findRepoRoot();
+    return path.join(repoRoot, '.continuum/sockets/continuum-core.sock');
+  }


getCoreSocketPath() builds the socket path relative to the repo root ("/.continuum/sockets/continuum-core.sock"), but the rest of the codebase resolves the IPC socket from the generated SOCKETS/SystemPaths (typically "$HOME/.continuum/sockets/continuum-core.sock" via getContinuumCoreSocketPath()). Spawning continuum-core-server with a different socket path will make RustCoreIPCClient/daemons connect to the wrong location and fail IPC. Please derive the socket path from the same single source of truth (e.g., import and use getContinuumCoreSocketPath()/SOCKETS.CONTINUUM_CORE, with the same absolute/relative resolution semantics) instead of hardcoding a repo-root-relative path.

Copilot · 2026-05-01T15:21:46Z

+    // audit log + dispatched to gist). Use that as the resolved-channel
+    // signal — params.channel is what WE asked for; this is what airc
+    // actually used after auto-scoping.
+    const resolvedChannel = this.parseResolvedChannel(stdout) ?? params.channel ?? '';


resolvedChannel fallback is ineffective: parseResolvedChannel() returns '' on no match, and the code uses nullish coalescing (??), so it will keep '' and never fall back to params.channel. This can cause channel to be reported as empty even when the caller provided --channel. Consider returning null/undefined from parseResolvedChannel when no match, or using a falsy check (||) for the fallback.

Suggested change

const resolvedChannel = this.parseResolvedChannel(stdout) ?? params.channel ?? '';

const resolvedChannel = this.parseResolvedChannel(stdout) || params.channel || '';

Copilot · 2026-05-01T15:21:46Z

  [SYSTEM_MILESTONES.SERVER_BOOTSTRAP_COMPLETE]: [SYSTEM_MILESTONES.SERVER_START],
  [SYSTEM_MILESTONES.SERVER_COMMANDS_LOADED]: [SYSTEM_MILESTONES.SERVER_START],
-  [SYSTEM_MILESTONES.SERVER_READY]: [SYSTEM_MILESTONES.SERVER_START],
+  [SYSTEM_MILESTONES.SERVER_READY]: [SYSTEM_MILESTONES.SERVER_START, SYSTEM_MILESTONES.CORE_READY],


SERVER_READY now depends on CORE_READY, but SystemOrchestrator.getCurrentState() still pre-marks SERVER_READY as completed based on the ready-signal/port checks without verifying the core socket. Because calculateMissingMilestones() short-circuits when a milestone is already in currentState, this can skip CORE_START/CORE_READY entirely on hot-restart / stale-signal scenarios (even if continuum-core-server is down). Either include CORE_READY in the "already completed" set only when the socket probe passes, or avoid marking SERVER_READY completed unless CORE_READY is also confirmed.

Copilot · 2026-05-01T15:21:46Z

+### Broadcast to the auto-scoped project room
+
+```bash
+undefined
+```
+
+### Broadcast to #general explicitly
+
+```bash
+undefined
+```
+
+### DM a specific peer
+
+```bash
+undefined
+```


The README examples render as literal undefined in bash code blocks, which is misleading for users trying to copy/paste. Please replace these with real CLI examples (e.g., ./jtag airc/send --message="..." and variants with --channel / --peer) or adjust the README generator to emit concrete example commands from the spec.

…egression + #978 nullish-coalescing cleanup THREE related changes from a live `npm start` test session 2026-05-01: 1. ALPHA-GAP-ANALYSIS.md is now THE single source of truth - Refreshed to 2026-05-01 with live-verified state - New "Today's Snapshot" section: what worked + broke in real `npm start` from feat/airc-send-command (#977 + #978 + #979 stack) - 3 new live-observed bugs in Phase 0: · NEW-A: continuum-core-server SIGABRT in vendored llama.cpp Metal `llm_build_smallthinker` cleanup. Real stack captured. · NEW-B: seed retries 21x/480s before giving up (concrete fail-fast fix designed) · NEW-C: shared/config.ts has /Users/joelteply/... HARDCODED (Carl-blocker) - 10 closed-since-Apr-17 items marked DONE - 21 new high-numbered open issues catalogued - Shortest path to "Install. Talk to AI." spelled out - Open PRs (continuum #976 #977 #978 #979 + airc #387) listed - Workflow note per Joel 2026-05-01: merge-to-canary, not PR-and-wait - Two predecessor docs DELETED + content folded: · docs/PRE-ALPHA-GAP-ANALYSIS.md (predates DMR pivot) · docs/planning/CARL-AND-DEV-PATH-TO-WORKING.md (interim) 2. SystemMilestones.ts — fix the #977 regression Original #977 added CORE_READY as SERVER_READY dep; consequence was browser never opens when Rust core SIGABRTs (Joel observed: "I don't see a browser"). This commit decouples them — SERVER_READY depends only on SERVER_START. SYSTEM_HEALTHY (monitoring signal) still requires both. Live-verified: browser opens despite SIGABRT-looping core. Joel confirmed: "opened good job." 3. AiLocalInference{Start,Status}ServerCommand.ts — || → ?? Three nullish-coalescing fixes left uncommitted from PR #978. NEXT STEPS for the test devices Joel just mentioned: 1. Verify NEW-C path bug repros on fresh test device (it should) 2. File NEW-A + NEW-C as GitHub issues 3. Trace seed-time llm_build_smallthinker call chain — likely a Candle-on-chat-hot-path bug per PR891 pivot 4. Implement seed fail-fast (~30 LOC) so install UX doesn't rot 8 minutes per attempt Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

… T7) Live-observed 2026-05-01 from M5 QA-Watcher tab Task 7: $ ./jtag airc/send --message="..." → stderr: "ERROR: Not initialized (/Users/joelteply/Development/cambrian/continuum/src/.airc). Run: airc connect" Root cause: spawn('airc', argv) inherited the daemon's CWD (typically src/ when invoked via ./jtag). airc's auto-scope rule walks up looking for a .airc/ — found nothing because src/.airc/ doesn't exist; the actual scope is at repo-root .airc/. Fix: belt-and-suspenders so the spawn is unambiguous about which scope it targets: - cwd: <repoRoot> → airc auto-scopes from continuum's git remote (→ #cambriantech), which IS the desired project-room behavior - env: AIRC_HOME=<repoRoot>/.airc → even if airc's CWD-walk were blocked or modified, AIRC_HOME pins the scope explicitly Added private static findRepoRoot() — walks up from CWD looking for .git or package.json with name='continuum'. Mirror of the same method in SystemOrchestrator (#977). Compression-deferred: when a 2nd airc-CLI-wrapping command lands (airc/peers, airc/whois, airc/identity/set), extract a BaseAircCommand with this helper as a protected method per the file header note. Verified: tsc --noEmit clean. End-to-end repro of the BUG was the M5-QA Task 7 broadcast that landed in airc #general (timestamp 2026-05-01T17:03:51Z). Composes with PR #979 — same outbox feature, different bug surface. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…r session Live-observed during the chat-with-AIs test session (Joel "you guys need to all remember to chat with the ais"): F1 (= existing #75 task): personas reply but with IDENTICAL canned text regardless of message content. Sent specific questions; got generic "Hello! I'm here to assist with any code review and analysis tasks..." back from multiple personas, recursive replies-to-replies. Cognition pipeline isn't engaging the message — generic-greeting template fires. THIS is the reason "AI doesn't really talk." F2 (NEW): ai/local-inference/start reports running:false after core SIGKILL+respawn. The Anthropic-compat HTTP server is initialized once via OnceCell at core startup; not re-triggered when core restarts. External agents pointing ANTHROPIC_BASE_URL would silently break on any core restart. Important for AGENT-BACKBONE Phase 1 reliability. F4 (NEW, CRITICAL): TS daemon's IPC client pool unrecoverable after core SIGKILL+respawn. ./jtag ping HANGS, ./jtag chat/send TIMES OUT. Sockets exist + accept connections + new core is alive, but commands don't complete. Full npm stop+start required to recover. THIS IS THE CARL-KILLER — every NEW-A SIGABRT in the wild puts users in this state. F4 supersedes the "#977 closes #722" claim. #977 Layer B (unlimited reconnect) gets the SOCKET back but the REQUEST PIPELINE is wedged. Three fix paths proposed in the doc: 1. Drain pending requests with "core restarted, reissue" error before reconnecting (so callers can retry) 2. Refuse new requests until pool cleanly drained 3. Re-create entire pool on detected core restart Composes with Task 8 supervisor-doesn't-own-pre-existing-cores: even when supervisor adopts an inherited core, IPC layer needs to handle "core changed under us" event. F4 is true regardless of who spawned the core. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…ia PID watcher M5-QA Task 8 (live-observed 2026-05-01) caught this: $ pgrep -x continuum-core-server # PID 67115 (alive 1h24m) $ kill -9 67115 # simulate SIGABRT $ sleep 30 $ pgrep -x continuum-core-server # NONE — supervisor never respawned Root cause: when parallel-start.sh's Phase 3 spawn beats orchestrator's executeCoreStart to it, executeCoreStart's isCoreSocketAlive() check correctly detects the existing core + skips the spawn. But this means this.coreProcess stays null + no on('exit') handler is attached. When the inherited core dies (NEW-A SIGABRT, kill -9, anything), the supervisor is BLIND to the death → no respawn. The original #977 design assumed the orchestrator OWNED the spawn. parallel-start.sh independently spawning continuum-core-server (since it predates this PR) breaks that assumption. THIS FIX (Task 8 layer): When isCoreSocketAlive=true at orchestrator start, attach a PID-poll watcher (`process.kill(pid, 0)` every 2s) on the inherited core's PID. When the watcher detects the PID is gone, spawnCoreProcess() is called to bring up a managed replacement — and from that point on, the normal on('exit') handler from spawnCoreProcess takes over the lifecycle. So the lifecycle transitions are: parallel-start.sh spawns core → orchestrator finds it via socket-alive → adoptInheritedCore registers PID-poll → inherited core dies (SIGABRT/kill) → watcher fires + spawnCoreProcess() → managed replacement now in this.coreProcess → normal supervisor path takes over API additions: - State: adoptedCorePid (number|null), adoptedCoreWatcher (interval handle) - Constant: ADOPTED_CORE_POLL_MS = 2_000 - Method: adoptInheritedCore(corePath, socketPath) - Method: findCoreProcessPid() — pgrep -x continuum-core-server - Method: stopAdoptedCoreWatcher() — idempotent cleanup - cleanup() now stops the adopted-core watcher first Failure-loud surface: if findCoreProcessPid() returns 0 (pgrep can't find it OR doesn't exist), we log a warn explaining the supervisor will be blind to the inherited core's death + return without crashing. Same intent as the never-swallow-errors rule — the gap is real, we surface it rather than pretend. What this STILL doesn't fix (separate scope): F4 (the carl-killer): TS daemon's IPC client pool can't recover even when supervisor respawns the core. Sockets reconnect but the request pipeline stays wedged. Fix is in ORMRustClient.ts (drain pending + reissue, OR refuse new until drained, OR recreate pool). Tracked in gap analysis under F4. F2 (local-inference HTTP server doesn't re-bind on core restart): when a managed replacement spawns, ai/local-inference/start needs to be re-triggered. Hooked off this fix's spawn callback in a follow-up. VALIDATION: - tsc --noEmit clean across the repo - Live deploy-test deferred since system is currently wedged from the SIGKILL test that surfaced T8 in the first place; will validate after npm stop+start (which the dev tab can trigger when ready) Composes with #977's existing supervisor + the dep-graph fix from ecb0eed. Closes part of #722 + the M5-QA T8 finding. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Test and others added 3 commits April 30, 2026 13:58

Copilot AI review requested due to automatic review settings May 1, 2026 15:10

Copilot started reviewing on behalf of joelteply May 1, 2026 15:11 View session

github-actions Bot added the size: XL label May 1, 2026

Copilot AI reviewed May 1, 2026

View reviewed changes

joelteply changed the base branch from main to canary May 1, 2026 16:15

Test and others added 3 commits May 1, 2026 13:53

joelteply mentioned this pull request May 1, 2026

fix(#974, Phase A of #981): self-aware required-status-checks for non-Docker PRs #982

Merged

joelteply merged commit c393c15 into canary May 1, 2026
5 of 7 checks passed

joelteply deleted the feat/airc-send-command branch May 1, 2026 20:19

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(airc/send): first-class command wrapping for persona outbox + dev-tooling#979

feat(airc/send): first-class command wrapping for persona outbox + dev-tooling#979
joelteply merged 7 commits into
canaryfrom
feat/airc-send-command

joelteply commented May 1, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI May 1, 2026

Uh oh!

Copilot AI May 1, 2026

Uh oh!

Copilot AI May 1, 2026

Uh oh!

Copilot AI May 1, 2026

Uh oh!

Copilot AI May 1, 2026

Uh oh!

Copilot AI May 1, 2026

Uh oh!

Copilot AI May 1, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

-undefined
+./jtag ai/local-inference/start
+# Returns:
+# {
+#   "url": "http://127.0.0.1:8421",
+#   "port": 8421,
+#   "protocol": "anthropic",
+#   "alreadyRunning": true
+# }

	const resolvedChannel = this.parseResolvedChannel(stdout) ?? params.channel ?? '';
	const resolvedChannel = this.parseResolvedChannel(stdout) \|\| params.channel \|\| '';

Conversation

joelteply commented May 1, 2026

Summary

What's added

Wire behavior

Deliberately not in v0

Compression-deferred notes (for future-me)

Validation

Composes with

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI May 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI May 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI May 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI May 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI May 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI May 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI May 1, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants