wicked-brain 0.16.0 → 0.17.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wicked-brain",
-  "version": "0.16.0",
+  "version": "0.17.0",
   "type": "module",
   "description": "Digital brain as skills for AI coding CLIs — no vector DB, no embeddings, no infrastructure",
   "keywords": [

package/server/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wicked-brain-server",
-  "version": "0.16.0",
+  "version": "0.17.0",
   "type": "module",
   "description": "SQLite FTS5 search server for wicked-brain digital knowledge bases",
   "keywords": [

package/skills/wicked-brain-ingest/SKILL.md

@@ -115,6 +115,7 @@ entities:
   people: [{people/roles}]
   programs: [{programs/initiatives}]
   metrics: ["{metric}: {value}"]
+method: {extraction method — see "Extraction method" below}
 confidence: {0.7 for text, 0.85 for vision}
 indexed_at: {current ISO timestamp}
 narrative_theme: {the "so what" in 8 words or fewer}
@@ -122,6 +123,23 @@ narrative_theme: {the "so what" in 8 words or fewer}
 
 {Extracted content in markdown format}
 
+## Extraction method
+
+The `method:` field records *how* the chunk's content was obtained — the
+provenance answer to "how do we know this?" It is distinct from `source_type`
+(which is the file format, e.g. `pdf`/`md`/`js`). Set it deterministically
+from the path you are already taking:
+
+- `deterministic-parse` — the TEXT path above (Read + split, no model judgement).
+- `llm-vision` — the BINARY path above (content extracted by the model viewing
+  the document/image).
+
+Use one of the controlled values: `deterministic-parse`, `llm-vision`,
+`llm-synthesis` (model-generated/inferred content), or `manual` (hand-authored).
+The value is plain frontmatter — it is stored and returned verbatim by the
+server with no schema migration. If omitted, downstream lint treats the chunk
+as `method: unknown`; prefer to set it explicitly.
+
 ## Tag Expansion
 
 After generating the initial `contains:` tags, expand each keyword with 1-3 synonyms or related terms:
@@ -334,6 +352,10 @@ async function ingestFile(filePath) {
       "  - text",
       "contains:",
       ...keywords.map(k => `  - ${k}`),
+      // method = HOW this chunk was obtained (provenance), distinct from
+      // source_type (file format). The batch path is a deterministic
+      // Read + split with no model judgement.
+      `method: deterministic-parse`,
       `confidence: 0.7`,
       `indexed_at: "${ts}"`,
       "---",

package/skills/wicked-brain-lint/SKILL.md

@@ -82,6 +82,21 @@ For each wiki article with source_hashes in frontmatter:
 ### Missing frontmatter
 Check each chunk has required frontmatter fields (source, chunk_id, confidence, indexed_at).
 
+Also check the **provenance** field `method` (how the chunk/memory was obtained:
+`deterministic-parse`, `llm-vision`, `llm-synthesis`, `manual`, or
+`session-capture` for memories). `method` is **optional** — it was added after
+some content was written, so a chunk/memory without it is still valid. When it
+is missing, auto-fix by stamping `method: unknown` and report the fix as `info`
+severity, type `missing_field` (do NOT raise it to a warning/error — that would
+invalidate pre-existing content). Surfacing `method: unknown` lets a reviewer
+distinguish facts with known provenance from those whose origin was never
+recorded.
+
+Lightweight provenance check (the "no source ⇒ assumption" rule): if a chunk has
+no `source`/`source_path` and its `method` is not one of the inferred kinds
+(`llm-synthesis`, `unknown`), flag it `info`, type `missing_field`:
+`unsourced fact with method "{method}" — add a source or set method to llm-synthesis`.
+
 ### Tag synonym candidates
 
 Call the server to get all tag frequencies:

package/skills/wicked-brain-memory/SKILL.md

@@ -100,6 +100,7 @@ Write to `{brain_path}/memory/{safe_name}.md`:
 ---
 type: {detected or provided type}
 tier: {resolved tier from Step 2b}
+method: {extraction method — see "Extraction method" below}
 confidence: 0.5
 importance: {from type defaults or override}
 ttl_days: {from type defaults or override, null if permanent}
@@ -117,6 +118,21 @@ indexed_at: "{ISO 8601 timestamp}"
 {memory content}
 ```
 
+#### Extraction method
+
+The `method:` field records *how* the memory was obtained — the provenance
+answer to "how do we know this?", mirroring the `method:` field on ingested
+chunks. Set it from how the memory came to be:
+
+- `session-capture` — captured live from the current session (the default for
+  "remember this" during work).
+- `manual` — explicitly stated by the user ("we decided X", interview-style).
+- `llm-synthesis` — inferred/derived by the agent rather than directly observed.
+
+Default to `session-capture` when unsure. The value is plain frontmatter,
+stored and returned verbatim by the server (no schema migration). If omitted,
+lint treats the memory as `method: unknown` — prefer to set it explicitly.
+
 #### Tier definitions
 
 - **working**: Active, session-specific context. Expires quickly (hours to days). Use for in-progress decisions, temporary notes, and things only relevant to the current task.
@@ -131,6 +147,7 @@ New memories start at the tier resolved from importance (default `episodic` for
 ---
 type: decision
 tier: semantic
+method: manual
 confidence: 0.9
 importance: 7
 ttl_days: null

package/skills/wicked-brain-synonyms/SKILL.md

@@ -43,8 +43,10 @@ Format:
 ```
 
 Keys are the short/common form. Values are expansions to try when the key
-appears in a search query. The search skill reads this file before executing
-queries and automatically expands sparse results using these mappings.
+appears in a search query. The default search path does NOT read this file —
+`wicked-brain:search` only loads it as a fallback when a direct search returns
+sparse results (0–2 matches), then re-runs the query with matching synonym
+values OR'd in and merges the results.
 
 ## Commands