Podcast ingest pipeline + RRF/MMR/SKOS retrieval stack + kernel perfo…

.gitignore

@@ -0,0 +1,79 @@
+# --- Operator-only docs / machine-local config ---
+CLAUDE.md
+.mcp.json
+.aiwg/
+.claude/
+.transcripts/
+
+# --- Secrets / credentials ---
+project/data/secrets.enc
+project/data/secrets.enc.tmp
+project/secrets/
+project/config.json
+project/config.json.bak
+
+# --- Model weights / large checkpoints ---
+project/.checkpoints/
+
+# --- Runtime caches and embeddings ---
+project/_embedding_cache.safetensors
+project/_embedding_cache_meta.json
+project/_embedding_cache.json
+project/embeddings/
+project/agent/embeddings/
+project/sdol/embeddings/
+
+# --- Caches and venv ---
+project/venv/
+project/__pycache__/
+**/__pycache__/
+*.pyc
+*.pyo
+
+# --- Install stamp files ---
+.deps_v*
+.datascience_v*
+.imagebind_v*
+.chatterbox_patch_v*
+.whisperx_v*
+
+# --- Runtime outputs (per leOS git etiquette) ---
+project/data/
+# Exception: SKOS concept hierarchy is seed config, not runtime state.
+!project/data/skos_hierarchy.json
+project/knowledgebase/
+project/media_library/files/
+project/media_library/derived/
+project/media_library/records/
+project/media_library/thumbnails/
+project/voice/data/
+project/projects/
+project/agent/data/
+project/bots/data/
+project/subsystems/evaluations/
+project/_media-processing/
+project/infra/decision_trees/audit_log.jsonl
+project/_system-maintenance/*.log
+project/_system-maintenance/*.err
+project/_system-maintenance/*_report.json
+project/_system-maintenance/*_report.csv
+project/_system-maintenance/files/
+project/_system-maintenance/history/
+project/_system-maintenance/postmortems/
+project/_system-maintenance/skills/
+project/_system-maintenance/project.json
+
+# --- Topic-centroid cache backups (regenerable from SEED_TOPICS) ---
+project/data/topic_centroids.json
+project/data/topic_centroids.json.old
+project/data/topic_centroids.json.before-*
+
+# --- Editor / OS noise ---
+.vscode/
+.idea/
+.DS_Store
+Thumbs.db
+*.swp
+
+# --- ComfyUI bundled install (run.bat clones on first boot) ---
+project/external/

project/README.md

@@ -707,8 +707,8 @@ work at every tier transition.
 
 Before any classified user query reaches the LLM, the substrate-
 gather package (`bots/substrate_gather/`) runs first.  This is the
-mechanism that turns "what is Xavier" from a 30-second LLM round trip
-into a 100ms KB lookup when an article on Xavier already exists.
+mechanism that turns "what is X" from a 30-second LLM round trip
+into a 100ms KB lookup when an article on X already exists.
 
 The companion brain classifies queries into six categories:
 **status_check**, **codebase_question**, **system_question**,

project/__init__.py

@@ -403,10 +403,10 @@ def _format_dict_brief(d, max_keys=8, depth=0, max_depth=1):
     runaway nesting on deeply-structured API payloads.
 
     For dexscreener-style data this means {baseToken: {symbol:
-    XAVIER, name: Xavier}} renders as:
+    ALPHA, name: Alpha}} renders as:
         - baseToken:
-            - symbol: XAVIER
-            - name: Xavier
+            - symbol: ALPHA
+            - name: Alpha
     instead of "baseToken: (dict, 2 items)".
     """
     if not d: