docs: add module location and cross-references to AI structuring modules

- mneme/core/prompts/json_parser.py: Location, related modules, tests
- mneme/core/ai_provider.py: Architecture hierarchy, related modules
- mneme/core/worker.py: StructuringWorker architecture, related modules
- mneme/core/heuristic_structuring.py: Method overview, related modules
- Bump version to 2.0.24

Author: barrelc

mneme/__init__.py

@@ -1,3 +1,3 @@
 """kimi-mneme — Persistent memory for Kimi Code CLI."""
 
-__version__ = "2.0.23"
+__version__ = "2.0.24"

mneme/core/ai_provider.py

@@ -1,4 +1,25 @@
-"""AI provider for observation structuring using configurable LLM backends."""
+"""AI provider for observation structuring using configurable LLM backends.
+
+Location: mneme/core/ai_provider.py
+
+This module provides the AI/Hybrid providers that orchestrate observation structuring.
+It sends tool call data to an LLM and parses the JSON response into structured observations.
+
+Architecture:
+    AIProvider (abstract)
+    ├── ConfigurableAIProvider  -- Uses any LLM client (Kimi, Ollama, OpenAI-compatible)
+    └── HybridProvider          -- Tries AI first, falls back to heuristic structuring
+
+Related modules:
+    - mneme/core/prompts/json_parser.py         -- Parses LLM JSON responses
+    - mneme/core/prompts/observation_prompt.py  -- Prompt templates
+    - mneme/core/heuristic_structuring.py       -- Fallback when AI is unavailable
+    - mneme/core/llm_client.py                  -- LLM client abstraction
+    - mneme/core/worker.py                      -- Background worker using HybridProvider
+    - mneme/db/structured_store.py              -- Stores ParsedObservation results
+
+Tests: tests/test_ai_provider.py
+"""
 
 from __future__ import annotations
 

mneme/core/heuristic_structuring.py

@@ -1,8 +1,29 @@
 """Heuristic structuring as fallback when AI is unavailable.
 
+Location: mneme/core/heuristic_structuring.py
+
 This module extracts structured observations from raw tool calls using
 rule-based analysis. It's designed to be fast, deterministic, and produce
 high-quality results comparable to AI structuring for common patterns.
+
+Architecture:
+    HeuristicStructuring
+    ├── structure()           -- Main entry point
+    ├── _detect_type()        -- Detects observation type from tool/error/prompt
+    ├── _generate_title()     -- Generates human-readable title
+    ├── _generate_subtitle()  -- Generates subtitle/summary
+    ├── _extract_facts()      -- Extracts key facts (up to 5)
+    ├── _generate_narrative() -- Generates narrative description
+    ├── _detect_concepts()    -- Detects concept tags from patterns
+    ├── _extract_files_*()    -- Extracts file paths
+    └── _extract_file_paths() -- Regex-based file path extraction
+
+Related modules:
+    - mneme/core/ai_provider.py         -- HybridProvider falls back to this module
+    - mneme/core/prompts/json_parser.py -- ParsedObservation dataclass (same fields)
+    - mneme/db/structured_store.py      -- Stores HeuristicObservation results
+
+Tests: tests/test_ai_provider.py (TestHeuristicStructuring)
 """
 
 from __future__ import annotations

mneme/core/prompts/json_parser.py

@@ -1,4 +1,19 @@
-"""Parse JSON responses from AI structuring."""
+"""Parse JSON responses from AI structuring.
+
+Location: mneme/core/prompts/json_parser.py
+
+This module is part of the AI structuring pipeline. It parses raw JSON responses
+from LLMs into structured ParsedObservation objects.
+
+Related modules:
+    - mneme/core/ai_provider.py        -- Uses parse_observation_json() via HybridProvider / ConfigurableAIProvider
+    - mneme/core/prompts/observation_prompt.py  -- Prompt templates sent to LLM
+    - mneme/core/heuristic_structuring.py       -- Fallback when AI is disabled
+    - mneme/core/worker.py             -- Background worker that orchestrates structuring
+    - mneme/db/structured_store.py     -- Stores the resulting ParsedObservation
+
+Tests: tests/test_json_parser.py
+"""
 
 from __future__ import annotations
 

mneme/core/worker.py

@@ -1,4 +1,27 @@
-"""Background worker for AI structuring queue."""
+"""Background worker for AI structuring queue.
+
+Location: mneme/core/worker.py
+
+This module runs a background loop that fetches pending observations from the queue,
+processes them through AI/heuristic structuring, and stores structured results.
+
+Architecture:
+    StructuringWorker
+    ├── _build_ai_provider_from_config()  -- Creates ConfigurableAIProvider from config
+    ├── start()                           -- Main loop (async)
+    ├── _process_batch()                  -- Claims messages, structures, stores results
+    └── _extract_project()                -- Extracts project name from cwd
+
+Related modules:
+    - mneme/core/ai_provider.py            -- HybridProvider / ConfigurableAIProvider
+    - mneme/core/prompts/json_parser.py    -- Parses LLM JSON responses
+    - mneme/core/heuristic_structuring.py  -- Fallback when AI is unavailable
+    - mneme/db/store.py                    -- Raw observation queue (claim_pending_messages)
+    - mneme/db/structured_store.py         -- Stores structured observations
+    - mneme/db/vector.py                   -- Vector search index
+
+Tests: tests/test_worker.py
+"""
 
 from __future__ import annotations
 

plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "kimi-mneme",
-  "version": "2.0.23",
+  "version": "2.0.24",
   "description": "Persistent memory plugin for Kimi Code CLI \u2014 search and retrieve past session context. Part of the kimi-plugins ecosystem.",
   "config_file": "config.json",
   "inject": {
@@ -111,4 +111,4 @@
       }
     }
   ]
-}
+}

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "kimi-mneme"
-version = "2.0.23"
+version = "2.0.24"
 description = "Persistent memory plugin for Kimi Code CLI"
 readme = "README.md"
 license = { text = "AGPL-3.0" }