P3.5-P3.7: 2-layer inventory, capability routing, STT/TTS adapters, Dev Contract

NCS:
- _collect_worker_caps() fetches capability flags from node-worker /caps
- _derive_capabilities() merges served model types + worker provider flags
- installed_artifacts replaces inventory_only (disk scan with DISK_SCAN_PATHS env)
- New endpoints: /capabilities/caps, /capabilities/installed

Node Worker:
- STT_PROVIDER, TTS_PROVIDER, OCR_PROVIDER, IMAGE_PROVIDER env flags
- /caps endpoint returns capabilities + providers for NCS aggregation
- STT adapter (providers/stt_mlx_whisper.py) — remote + local mode
- TTS adapter (providers/tts_mlx_kokoro.py) — remote + local mode
- OCR handler via vision_prompted (ollama_vision with OCR prompt)
- NATS subjects: node.{id}.stt/tts/ocr/image.request

Router:
- POST /v1/capability/{stt,tts,ocr,image} — capability-based offload routing
- GET /v1/capabilities — global view with capabilities_by_node
- require_fresh_caps(ttl) preflight guard
- find_nodes_with_capability(cap) + load-based node selection

Ops:
- ops/fabric_snapshot.py — full runtime snapshot collector
- ops/fabric_preflight.sh — quick check + snapshot save + diff
- docs/fabric_contract.md — Dev Contract v0.1 (preflight-first)
- tests/test_fabric_contract.py — CI enforcement (6 tests)

Made-with: Cursor

services/node-worker/worker.py

@@ -10,6 +10,7 @@ import config
 from models import JobRequest, JobResponse, JobError
 from idempotency import IdempotencyStore
 from providers import ollama, ollama_vision
+from providers import stt_mlx_whisper, tts_mlx_kokoro
 import fabric_metrics as fm
 
 logger = logging.getLogger("node-worker")
@@ -34,6 +35,7 @@ async def start(nats_client):
         f"node.{nid}.stt.request",
         f"node.{nid}.tts.request",
         f"node.{nid}.image.request",
+        f"node.{nid}.ocr.request",
     ]
     for subj in subjects:
         await nats_client.subscribe(subj, cb=_handle_request)
@@ -175,14 +177,52 @@ async def _execute(job: JobRequest, remaining_ms: int) -> JobResponse:
                 ),
                 timeout=timeout_s,
             )
-        elif job.required_type in ("stt", "tts", "image"):
+        elif job.required_type == "stt":
+            if config.STT_PROVIDER == "none":
+                return JobResponse(
+                    job_id=job.job_id, trace_id=job.trace_id, node_id=config.NODE_ID,
+                    status="error",
+                    error=JobError(code="NOT_AVAILABLE", message="STT not configured on this node"),
+                )
+            result = await asyncio.wait_for(
+                stt_mlx_whisper.transcribe(payload), timeout=timeout_s,
+            )
+        elif job.required_type == "tts":
+            if config.TTS_PROVIDER == "none":
+                return JobResponse(
+                    job_id=job.job_id, trace_id=job.trace_id, node_id=config.NODE_ID,
+                    status="error",
+                    error=JobError(code="NOT_AVAILABLE", message="TTS not configured on this node"),
+                )
+            result = await asyncio.wait_for(
+                tts_mlx_kokoro.synthesize(payload), timeout=timeout_s,
+            )
+        elif job.required_type == "ocr":
+            if config.OCR_PROVIDER == "none":
+                return JobResponse(
+                    job_id=job.job_id, trace_id=job.trace_id, node_id=config.NODE_ID,
+                    status="error",
+                    error=JobError(code="NOT_AVAILABLE", message="OCR not configured on this node"),
+                )
+            ocr_prompt = payload.get("prompt", "Extract all text from this image. Return JSON: {\"text\": \"...\", \"language\": \"...\"}")
+            result = await asyncio.wait_for(
+                ollama_vision.infer(
+                    images=payload.get("images"),
+                    prompt=ocr_prompt,
+                    model=model or config.DEFAULT_VISION,
+                    system="You are an OCR engine. Extract text precisely. Return valid JSON only.",
+                    max_tokens=hints.get("max_tokens", 4096),
+                    temperature=0.05,
+                    timeout_s=timeout_s,
+                ),
+                timeout=timeout_s,
+            )
+            result["provider"] = "vision_prompted_ocr"
+        elif job.required_type == "image":
             return JobResponse(
                 job_id=job.job_id, trace_id=job.trace_id, node_id=config.NODE_ID,
                 status="error",
-                error=JobError(
-                    code="NOT_YET_IMPLEMENTED",
-                    message=f"{job.required_type} adapter coming soon; use direct runtime API for now",
-                ),
+                error=JobError(code="NOT_YET_IMPLEMENTED", message="Image adapter pending P3.7"),
             )
         else:
             return JobResponse(