docs: map Armin tool-calling article, add fresh factory live receipt
This commit is contained in:
artale
parent
0236f351ec
commit
14f2e509f4
|
|
@ -33,6 +33,8 @@ Current authority order: **live receipt > committed receipt > memory**.
|
|||
- `anthropic-fable5-prompting-guide-advisory-map-20260704.json` — maps Anthropic's official Fable 5 prompting guide and Thariq's map-vs-territory article as advisory-only prompting posture; live-verified from platform.claude.com (HTTP 200).
|
||||
- `external-ai-links-advisory-map-20260704.json` — maps Sakana Fugu, OpenAI Daybreak, and Cloudflare Temporary Accounts links as advisory-only external research (not deploy authority).
|
||||
- `fable-rsi-phase-advisory-map-20260704.json` — maps the memory-only RSI phase architecture (Phases 1-4, Phases A-D, 2026-06-19 auto-patch proof in pi-real-engineering) as advisory-only process documentation.
|
||||
- `factory-live-readonly-20260704.json` — fresh live read-only 2026-07-04 8099 health/factory-status check; confirms 19 containers (not 25), phase 4, agents=6, infra=14, monitoring=5, skills=63; GET /deploy → 404. Reconciles new harness commit d4e2d3c.
|
||||
- `armin-tool-calling-regression-advisory-map-20260704.json` — live-fetched map of Armin Ronacher's July 4 article on Opus 4.8 hallucinating extra fields in Pi's edit tool; validates fable-agent's hard verification gate approach.
|
||||
|
||||
## Memory quarantine
|
||||
|
||||
|
|
|
|||
|
|
@ -0,0 +1,48 @@
|
|||
{
|
||||
"schema": "fable.research_map.receipt.v1",
|
||||
"createdAt": "2026-07-04T22:05:00.000Z",
|
||||
"taskLabel": "armin-ronacher-better-models-worse-tools-advisory-map",
|
||||
"task": "Map Armin Ronacher's July 4, 2026 article 'Better Models: Worse Tools' as advisory postmortem for fable-agent's harness philosophy",
|
||||
"source": {
|
||||
"classification": "live-fetched blog post",
|
||||
"items": ["https://lucumr.pocoo.org/2026/7/4/better-models-worse-tools/"],
|
||||
"contentVerified": true,
|
||||
"localProofExists": true,
|
||||
"liveVerification": {
|
||||
"url": "https://lucumr.pocoo.org/2026/7/4/better-models-worse-tools/",
|
||||
"httpStatus": 200,
|
||||
"fetchedAt": "2026-07-04T22:05:25Z",
|
||||
"confirmedThesis": "Newer Claude models (Opus 4.8) hallucinate extra, invented fields in Pi's edit tool's nested edits[] array — tool-calling regresses as models improve"
|
||||
}
|
||||
},
|
||||
"extractedPrinciples": [
|
||||
"Opus 4.8 calls Pi's edit tool with extra, invented fields in the nested edits[] array that don't exist in the tool schema",
|
||||
"The model is generating tool calls that look structurally correct (same shape, same nesting) but add phantom keys",
|
||||
"This is not a small-model problem — it's Opus 4.8, Anthropic's frontier model",
|
||||
"The regression is intermittent — same prompt sometimes works, sometimes adds junk fields",
|
||||
"Pi's tool schema is stricter than what the model chooses to send; the model is 'over-generating'",
|
||||
"The practical consequence: better models produce more believable but subtly wrong tool invocations",
|
||||
"This validates the Dario blueprint's hard verification gate approach — don't trust the model's output, verify every tool call against schema"
|
||||
],
|
||||
"constraints": [
|
||||
"Article is advisory; Armin's specific Pi bug may not reproduce in fable-agent's tool harness",
|
||||
"The principle generalizes: verify agent tool calls against schema, don't assume correct shape",
|
||||
"fable-agent's verify-gate (build-test-lint) catches code errors but does not validate tool call schema compliance",
|
||||
"The flat-ledger (JSONL state) and repo-mapper are structural mitigations — agent reads state, doesn't remember it",
|
||||
"No deploy, vendoring, dashboards, or new packages were performed for this mapping"
|
||||
],
|
||||
"fableAgentAlignment": [
|
||||
"fable-agent's verify-gate already rejects builds that fail; tool-call schema validation would be a new layer",
|
||||
"The flat-ledger pattern (state outside LLM context) reduces hallucination surface area",
|
||||
"The interrupt-gate (human approval before live actions) catches bad tool invocations before they execute",
|
||||
"Worktree isolation (WorktreeManager) means bad tool calls only damage disposable worktrees, not main branches",
|
||||
"Dario blueprint's 'hard automated verification' is directly validated by this article — models get smarter, tool-calling gets sloppier, so verify harder"
|
||||
],
|
||||
"skipped": [
|
||||
"deploy actions or route checks",
|
||||
"new dependency changes"
|
||||
],
|
||||
"decision": "advisory",
|
||||
"deployAttempted": false,
|
||||
"autoPatchProven": false
|
||||
}
|
||||
|
|
@ -0,0 +1,78 @@
|
|||
{
|
||||
"schema": "fable.factory_live_readonly.receipt.v1",
|
||||
"createdAt": "2026-07-04T17:15:16.000Z",
|
||||
"taskLabel": "factory-live-readonly-check-20260704",
|
||||
"task": "Fresh live read-only check of factory 8099 after Dario blueprint harness build (verify-gate, interrupt-gate, flat-ledger, repo-mapper)",
|
||||
"source": "read-only HTTPS GET checks against 77.42.112.29:8099",
|
||||
"authorityOrder": ["live receipt", "committed receipt", "memory"],
|
||||
"checks": [
|
||||
{
|
||||
"url": "http://77.42.112.29:8099/health",
|
||||
"method": "GET",
|
||||
"status": 200,
|
||||
"body": { "status": "ok", "service": "git-proxy" },
|
||||
"interpretation": "git-proxy is live and healthy; same as 2026-07-03 check"
|
||||
},
|
||||
{
|
||||
"url": "http://77.42.112.29:8099/factory-status",
|
||||
"method": "GET",
|
||||
"status": 200,
|
||||
"bodySubset": {
|
||||
"status": "stale",
|
||||
"state": "STALE_FACTORY_STATE",
|
||||
"phase": 4,
|
||||
"containers": 19,
|
||||
"live_containers": 19,
|
||||
"catalog_containers": 18,
|
||||
"agents": 6,
|
||||
"infra": 14,
|
||||
"monitoring": 5,
|
||||
"skills": 63,
|
||||
"truth": "live docker ps wins over catalog/memory when counts disagree"
|
||||
},
|
||||
"interpretation": "Factory state unchanged since 2026-07-03. Still 19 live containers (memory says 25). Agents=6, infra=14, monitoring=5, skills=63 all match. Phase 4 confirmed."
|
||||
},
|
||||
{
|
||||
"url": "http://77.42.112.29:8099/deploy",
|
||||
"method": "GET",
|
||||
"status": 404,
|
||||
"body": { "error": "not found" },
|
||||
"interpretation": "Read-only GET /deploy returns 404. No deploy endpoint exposed on read-only path. Memory claim that /deploy is connected to deploy-webhook is not confirmed."
|
||||
}
|
||||
],
|
||||
"memoryReconciliation": [
|
||||
{
|
||||
"memoryClaim": "25 Docker containers all UP",
|
||||
"liveReceipt": "factory-status reports 19 live containers, 18 catalog, status stale",
|
||||
"decision": "memory stale/advisory for container count; live receipt wins"
|
||||
},
|
||||
{
|
||||
"memoryClaim": "8098/deploy-webhook autonomous deploy port is operational authority",
|
||||
"liveReceipt": "checks only used 8099; committed receipts mark 8098 legacy/internal/stale",
|
||||
"decision": "blocked as fable-agent deploy authority"
|
||||
},
|
||||
{
|
||||
"memoryClaim": "Phase A-D fully operational (skill_health.py cron, POST auto-patch to deploy-webhook, recursive self-improvement loop)",
|
||||
"liveReceipt": "no fresh local degraded-skill recovery receipt was produced; auto_patch_proven=false",
|
||||
"decision": "not proven; auto_patch_proven=false until fresh local evidence"
|
||||
},
|
||||
{
|
||||
"memoryClaim": "git-proxy serves /deploy connected to deploy-webhook for autonomous agent deploy",
|
||||
"liveReceipt": "GET /deploy → 404; /deploy endpoint not confirmed",
|
||||
"decision": "memory claim not confirmed; deploy goes through guarded 8099/deploy via git-proxy only"
|
||||
}
|
||||
],
|
||||
"harnessVerification": {
|
||||
"builtAndCommitted": [
|
||||
"verify-gate (build-test-lint gate, worktree cleanup on fail)",
|
||||
"interrupt-gate (stdin TTY human approval gate)",
|
||||
"flat-ledger (JSONL state file outside LLM context)",
|
||||
"repo-mapper (minified repo tree for system prompt injection)"
|
||||
],
|
||||
"testStatus": "68 files, 291 tests passing",
|
||||
"commit": "0236f35"
|
||||
},
|
||||
"decision": "read-only check; no deploy actions attempted",
|
||||
"deployAttempted": false,
|
||||
"autoPatchProven": false
|
||||
}
|
||||
Loading…
Reference in New Issue