NWcortex 1.0.2__tar.gz → 1.0.4__tar.gz

{nwcortex-1.0.2 → nwcortex-1.0.4}/LICENSE

@@ -1,21 +1,21 @@
-MIT License
-
-Copyright (c) 2025 Thatgfsj
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+MIT License
+
+Copyright (c) 2025 Thatgfsj
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

{nwcortex-1.0.2 → nwcortex-1.0.4}/MANIFEST.in

@@ -1,10 +1,11 @@
-include README.md
-include LICENSE
-include star_graph/defaults.yaml
-recursive-include examples *.py
-prune venv
-prune dist
-prune build
-prune .git
-prune __pycache__
-global-exclude *.pyc
+include README.md
+include LICENSE
+include star_graph/defaults.yaml
+recursive-include star_graph/contrib *.py
+recursive-include examples *.py
+prune venv
+prune dist
+prune build
+prune .git
+prune __pycache__
+global-exclude *.pyc

{nwcortex-1.0.2 → nwcortex-1.0.4/NWcortex.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: NWcortex
-Version: 1.0.2
+Version: 1.0.4
 Summary: NeuroWeave Cortex (NWC) — Graph-first cognitive memory runtime for AI agents. Hippocampal-inspired architecture with domain routing, spreading activation, edge budget management, 4-layer compression, thermal storage, personality modeling, and 8-phase sleep consolidation.
 Author-email: Thatgfsj <thatgfsj@users.noreply.github.com>
 License: MIT
@@ -24,6 +24,11 @@ Requires-Dist: PyYAML>=6.0
 Requires-Dist: scikit-learn>=1.3.0
 Provides-Extra: embeddings
 Requires-Dist: sentence-transformers>=2.2.0; extra == "embeddings"
+Provides-Extra: mcp
+Requires-Dist: mcp>=1.0.0; extra == "mcp"
+Provides-Extra: all
+Requires-Dist: sentence-transformers>=2.2.0; extra == "all"
+Requires-Dist: mcp>=1.0.0; extra == "all"
 Dynamic: license-file
 
 # NeuroWeave Cortex (NWC) — Cognitive Memory Runtime
@@ -46,6 +51,12 @@ Vector databases retrieve. Graph databases traverse. NeuroWeave Cortex runs a **
 | Automatic forgetting (survival decay) | no | no | yes |
 | Memory strengthening (rehearsal) | no | no | yes |
 | Conflict detection (contradiction edges) | no | no | yes |
+| Multi-strategy RRF retrieval fusion | no | no | yes |
+| Cross-Encoder reranking | no | no | yes |
+| Explainable reasoning paths | no | no | yes |
+| Memory revision engine (sleep) | no | no | yes |
+| Hot/Warm/Cold memory tiering | no | no | yes |
+| Multimodal (text+image+audio) | no | no | yes |
 | Fuzzy recall ("I think I remember...") | no | no | yes |
 | Emergent abstraction (pattern discovery) | no | no | yes |
 | Temporal context (TimeSpine-indexed) | no | no | yes |
@@ -92,6 +103,9 @@ print(report)
 # Persist
 mgr.save("agent_memory.json")
 mgr.load("agent_memory.json")
+
+# Export to editable Markdown (GBrain-aligned)
+mgr.export_markdown("memories/")
 ```
 
 ## Architecture
@@ -109,12 +123,15 @@ Layer 3: Behavior    │  Cortex routing, memory gating, working memory,
                      │
 Layer 2: Cognitive   │  Hub abstraction, cascade recall, TimeSpine temporal index,
     (retriever.py,   │  sleep consolidation, evolution, ghost revival,
-     sleep.py,       │  abstraction, community detection, competition
-     evolution.py,   │  "How do memories connect, strengthen, and fade?"
-     ghost.py,       │
+     sleep.py,       │  abstraction, community detection, competition,
+     evolution.py,   │  conflict detection, memory revision, cross-encoder
+     ghost.py,       │  "How do memories connect, strengthen, and fade?"
      abstraction.py, │
      community.py,   │
      competition.py, │
+     conflict_detection.py,│
+     memory_revision.py,   │
+     cross_encoder.py,     │
      timespine.py,   │
      cascade.py,     │
      hub.py)         │
@@ -139,11 +156,14 @@ Layer 1: Storage     │  CRUD, persistence, ANN indexing, tiered storage,
 | `retrieval_pipeline.py` | 5-layer dimensional descent (L0→L4) with automatic degradation |
 | `scheduler.py` | Context-aware retrieval with memory type selection |
 | `working_memory.py` | Short-term buffer (15-item, 1h TTL) — auto-promotes to long-term |
-| `sleep.py` | 8-phase sleep: N1_Replay → N2_Merge → N3_Compression → N3b_AtomFacts → REM_Emotion → N4_Prune → N5_HubConnect → N6_IndexRebuild |
+| `sleep.py` | 10-phase sleep: N1_Replay → N2_Merge → N2b_ConflictDetection → N2c_MemoryRevision → N3_Compression → N3d_SleepRebuild → REM_Emotion → N4_Prune (v1.0.4) |
 | `evolution.py` | Survival-based decay (Ebbinghaus/Power-law/Exponential), belief transitions, interference |
 | `retriever.py` | HybridFusion + OscillationResonance + VectorSimilarity + Personalized PageRank + explainable scores |
 | `dual_channel.py` | System-1 (fast) + System-2 (deep) dual-channel retrieval with auto-trigger |
-| `bm25.py` | Sparse keyword retrieval (BM25) with reciprocal rank fusion for hybrid search |
+| `bm25.py` | Sparse keyword retrieval (BM25) with weighted reciprocal rank fusion for hybrid search |
+| `cross_encoder.py` | Post-retrieval Cross-Encoder reranking (sentence-transformers) for precision improvement |
+| `conflict_detection.py` | Semantic contradiction detection with overwrite/coexist/deprecate resolution strategies |
+| `memory_revision.py` | Sleep-phase memory revision engine — surprise-prioritized, template/LLM re-summarization |
 | `ghost.py` | Latent memory traces with fuzzy recall and contradiction tracking (NegativeGhost) |
 | `abstraction.py` | Emergent category discovery from anchor clusters |
 | `community.py` | Louvain community detection with centroid routing |
@@ -162,6 +182,11 @@ Layer 1: Storage     │  CRUD, persistence, ANN indexing, tiered storage,
 | `resonance.py` | Phase-locked oscillation resonance for temporal-coherent retrieval |
 | `streaming.py` | Streaming memory buffer with backpressure |
 | `benchmark.py` | Built-in benchmark suite (5 categories) |
+| `batch_vectorizer.py` | Deferred batch embedding writes (buffer >= 32 / >30s flush) with SQLite WAL crash recovery |
+| `zero_llm_pipeline.py` | 7-stage zero-LLM ingestion: security → embed → dedup → entity → classify → score → link |
+| `multimodal.py` | Cross-modal text/image/audio memory — CLIP image encoding, Whisper/spectrogram audio, unified embedding space |
+| `forget_certificate.py` | Ed25519-signed JWS deletion certificates — GDPR Article 17 provable erasure |
+| `markdown_export.py` | Operator-owned plain-text memory export (GBrain-aligned); timeline + topic Markdown |
 | `config.py` | Centralized YAML config with schema validation, dot-path access, overrides |
 
 ## Retrieval Benchmarks
@@ -264,11 +289,14 @@ See `star_graph/defaults.yaml` for all 300+ tunable parameters.
 # From PyPI (package name: NWcortex, import as: star_graph)
 pip install NWcortex
 
-# With sentence-transformers for semantic embeddings (~2GB model download on first use)
+# With sentence-transformers for semantic embeddings
 pip install "NWcortex[embeddings]"
 
-# Optional: for SQLite storage backend
-pip install aiosqlite
+# With MCP server support (for AI agent integration)
+pip install "NWcortex[mcp]"
+
+# Everything
+pip install "NWcortex[all]"
 
 # Run demo
 python examples/emergence_demo.py
@@ -276,6 +304,91 @@ python examples/emergence_demo.py
 
 **Note:** Without `[embeddings]`, the system uses a lightweight TF-IDF fallback for text encoding. Install `sentence-transformers` only if you need semantic-quality embeddings.
 
+## MCP Server — AI Agent Integration
+
+NeuroWeave Cortex ships with a [Model Context Protocol](https://modelcontextprotocol.io) server. Connect any MCP-compatible AI agent (Claude Desktop, OpenClaw, Cursor, etc.) for persistent cognitive memory across conversations.
+
+### Quick start
+
+```bash
+# Install with MCP support
+pip install "NWcortex[mcp]"
+
+# Launch the MCP server on stdio
+nwc-mcp
+
+# With persistent storage
+nwc-mcp --storage agent_memory.json
+
+# Load a previously saved memory graph
+nwc-mcp --load my_memories.json
+```
+
+### Tool reference (12 tools)
+
+| Tool | Description |
+|------|-------------|
+| `remember` | Store a long-term memory (tags, importance, emotional valence) |
+| `remember_working` | Store in fast ephemeral working-memory buffer (current task context) |
+| `recall` | Context-aware semantic retrieval with task-type routing |
+| `forget` | Remove a memory, creating a ghost trace for potential fuzzy recall |
+| `sleep` | Run 5-phase sleep consolidation — merges, prunes, forms schemas |
+| `consolidate` | Micro-consolidation — incremental, non-blocking |
+| `stats` | Memory system statistics (anchors, edges, ghosts, schemas, cognitive health) |
+| `fuzzy_recall` | Low-confidence recall from ghost traces ("I seem to remember...") |
+| `get_profile` | Inferred user profile from accumulated memories |
+| `evolve` | Memory evolution cycle (decay, boost, conflict resolution) |
+| `save` | Persist memory graph to disk (JSON) |
+| `load` | Load memory graph from disk (JSON) |
+
+### Claude Desktop config
+
+```json
+{
+  "mcpServers": {
+    "neuroweave-cortex": {
+      "command": "nwc-mcp",
+      "args": ["--storage", "/path/to/agent_memory.json"]
+    }
+  }
+}
+```
+
+### OpenClaw config
+
+```json
+{
+  "mcpServers": {
+    "neuroweave-cortex": {
+      "command": "nwc-mcp",
+      "args": ["--storage", "/path/to/agent_memory.json"]
+    }
+  }
+}
+```
+
+## REST API Server
+
+```bash
+# Start the REST server
+nwc-server --port 8420
+
+# Or via module
+python -m star_graph.server --port 8420
+```
+
+Endpoints:
+
+| Method | Path | Description |
+|--------|------|-------------|
+| GET | `/health` | Health check with anchor/edge counts |
+| GET | `/metrics` | Prometheus-format metrics |
+| GET | `/stats` | Full memory system statistics |
+| POST | `/remember` | Store a memory `{"text": "...", "tags": [...]}` |
+| POST | `/recall` | Retrieve memories `{"query": "...", "max_items": 10}` |
+| POST | `/sleep` | Run sleep consolidation |
+| POST | `/consolidate` | Run micro-consolidation |
+
 ## Interactive Demo
 
 ```bash
@@ -287,6 +400,9 @@ python examples/emergence_demo.py
 # sg-query --trace "When did Alice visit Hawaii?"
 # sg-stats --schemas --ghosts
 # sg-sleep --retention 0.15 --edge-prune 0.1
+# nwc-export --output memories/ --organize both
+# nwc-forget <anchor_id> --certificate --reason gdpr_art17
+# nwc-verify certificates/forget-<id>.jws
 ```
 
 ## Benchmarks
@@ -306,7 +422,7 @@ pytest tests/ -v
 pytest tests/ --cov=star_graph --cov-report=term
 ```
 
-**Status:** 1,989 tests passing, 80% coverage (v1.0.0).
+**Status:** 1,989 tests passing, 84 modules (v1.0.4-dev).
 
 ## Roadmap
 

{nwcortex-1.0.2 → nwcortex-1.0.4}/NWcortex.egg-info/SOURCES.txt

@@ -8,6 +8,7 @@ NWcortex.egg-info/dependency_links.txt
 NWcortex.egg-info/entry_points.txt
 NWcortex.egg-info/requires.txt
 NWcortex.egg-info/top_level.txt
+benchmarks/run_benchmarks.py
 examples/ablation_benchmark.py
 examples/emergence_demo.py
 examples/locomo_eval.py
@@ -22,6 +23,7 @@ star_graph/anchor.py
 star_graph/async_manager.py
 star_graph/atom_facts.py
 star_graph/autobiography.py
+star_graph/batch_vectorizer.py
 star_graph/benchmark.py
 star_graph/bm25.py
 star_graph/brain_sphere.py
@@ -35,8 +37,10 @@ star_graph/competition.py
 star_graph/compiler.py
 star_graph/compression.py
 star_graph/config.py
+star_graph/conflict_detection.py
 star_graph/cortex.py
 star_graph/cost_estimator.py
+star_graph/cross_encoder.py
 star_graph/defaults.yaml
 star_graph/domain_router.py
 star_graph/dual_channel.py
@@ -47,6 +51,7 @@ star_graph/embedding.py
 star_graph/episodic_memory.py
 star_graph/evolution.py
 star_graph/exact_cache.py
+star_graph/forget_certificate.py
 star_graph/four_layer.py
 star_graph/gate.py
 star_graph/ghost.py
@@ -59,8 +64,10 @@ star_graph/layers.py
 star_graph/logger.py
 star_graph/manager.py
 star_graph/manager_stats.py
+star_graph/markdown_export.py
 star_graph/math_utils.py
 star_graph/mcp_server.py
+star_graph/memory_revision.py
 star_graph/metrics.py
 star_graph/micro_sleep.py
 star_graph/multimodal.py
@@ -105,7 +112,27 @@ star_graph/tracing.py
 star_graph/versioned_memory.py
 star_graph/working_memory.py
 star_graph/write_gate.py
+star_graph/zero_llm_pipeline.py
 star_graph/consolidation/__init__.py
+star_graph/consolidation/abstraction.py
+star_graph/consolidation/community.py
+star_graph/consolidation/competition.py
+star_graph/consolidation/compression.py
+star_graph/consolidation/conflict_detection.py
+star_graph/consolidation/domain_router.py
+star_graph/consolidation/evolution.py
+star_graph/consolidation/ghost.py
+star_graph/consolidation/hub.py
+star_graph/consolidation/memory_revision.py
+star_graph/consolidation/micro_sleep.py
+star_graph/consolidation/reflection_loop.py
+star_graph/consolidation/self_org.py
+star_graph/consolidation/sleep.py
+star_graph/consolidation/sleep_consolidate.py
+star_graph/consolidation/sleep_nrem.py
+star_graph/consolidation/sleep_rem.py
+star_graph/consolidation/sleep_report.py
+star_graph/consolidation/survival.py
 star_graph/contrib/__init__.py
 star_graph/contrib/benchmark.py
 star_graph/contrib/mcp_server.py
@@ -113,13 +140,43 @@ star_graph/contrib/snapshot.py
 star_graph/contrib/streaming.py
 star_graph/contrib/symbolic_filter.py
 star_graph/cortex_api/__init__.py
+star_graph/cortex_api/async_manager.py
+star_graph/cortex_api/gate.py
+star_graph/cortex_api/manager.py
+star_graph/cortex_api/manager_stats.py
+star_graph/cortex_api/mcp_server.py
+star_graph/cortex_api/runtime.py
+star_graph/cortex_api/runtime_core.py
+star_graph/cortex_api/runtime_lifecycle.py
+star_graph/cortex_api/runtime_stats.py
+star_graph/cortex_api/scheduler.py
+star_graph/cortex_api/server.py
+star_graph/cortex_api/working_memory.py
+star_graph/cortex_api/write_gate.py
 star_graph/embedding_provider/__init__.py
 star_graph/embedding_provider/providers.py
 star_graph/extras/__init__.py
 star_graph/extras/autobiography.py
 star_graph/extras/resonance.py
 star_graph/memory_core/__init__.py
+star_graph/memory_core/anchor.py
+star_graph/memory_core/graph.py
+star_graph/memory_core/index.py
+star_graph/memory_core/sqlite_storage.py
+star_graph/memory_core/storage.py
+star_graph/memory_core/storage_backend.py
+star_graph/memory_core/tier.py
+star_graph/memory_core/tiered.py
 star_graph/retrieval_engine/__init__.py
+star_graph/retrieval_engine/bm25.py
+star_graph/retrieval_engine/cognitive_cache.py
+star_graph/retrieval_engine/cross_encoder.py
+star_graph/retrieval_engine/dual_channel.py
+star_graph/retrieval_engine/exact_cache.py
+star_graph/retrieval_engine/retrieval_budget.py
+star_graph/retrieval_engine/retrieval_core.py
+star_graph/retrieval_engine/retrieval_pipeline.py
+star_graph/retrieval_engine/retriever.py
 tests/test_abstraction.py
 tests/test_abstractive_memory.py
 tests/test_anchor.py

nwcortex-1.0.4/NWcortex.egg-info/entry_points.txt

@@ -0,0 +1,10 @@
+[console_scripts]
+nwc-export = star_graph.cli:export_cmd
+nwc-forget = star_graph.cli:forget_cmd
+nwc-mcp = star_graph.mcp_server:entry
+nwc-server = star_graph.server:entry
+nwc-verify = star_graph.cli:verify_cmd
+sg-add = star_graph.cli:add
+sg-query = star_graph.cli:query
+sg-sleep = star_graph.cli:sleep_cmd
+sg-stats = star_graph.cli:stats_cmd

{nwcortex-1.0.2 → nwcortex-1.0.4}/NWcortex.egg-info/requires.txt

@@ -2,5 +2,12 @@ numpy>=1.24.0
 PyYAML>=6.0
 scikit-learn>=1.3.0
 
+[all]
+sentence-transformers>=2.2.0
+mcp>=1.0.0
+
 [embeddings]
 sentence-transformers>=2.2.0
+
+[mcp]
+mcp>=1.0.0

nwcortex-1.0.4/NWcortex.egg-info/top_level.txt

@@ -0,0 +1,4 @@
+benchmarks
+certificates
+openclaw-plugin
+star_graph

{nwcortex-1.0.2/NWcortex.egg-info → nwcortex-1.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: NWcortex
-Version: 1.0.2
+Version: 1.0.4
 Summary: NeuroWeave Cortex (NWC) — Graph-first cognitive memory runtime for AI agents. Hippocampal-inspired architecture with domain routing, spreading activation, edge budget management, 4-layer compression, thermal storage, personality modeling, and 8-phase sleep consolidation.
 Author-email: Thatgfsj <thatgfsj@users.noreply.github.com>
 License: MIT
@@ -24,6 +24,11 @@ Requires-Dist: PyYAML>=6.0
 Requires-Dist: scikit-learn>=1.3.0
 Provides-Extra: embeddings
 Requires-Dist: sentence-transformers>=2.2.0; extra == "embeddings"
+Provides-Extra: mcp
+Requires-Dist: mcp>=1.0.0; extra == "mcp"
+Provides-Extra: all
+Requires-Dist: sentence-transformers>=2.2.0; extra == "all"
+Requires-Dist: mcp>=1.0.0; extra == "all"
 Dynamic: license-file
 
 # NeuroWeave Cortex (NWC) — Cognitive Memory Runtime
@@ -46,6 +51,12 @@ Vector databases retrieve. Graph databases traverse. NeuroWeave Cortex runs a **
 | Automatic forgetting (survival decay) | no | no | yes |
 | Memory strengthening (rehearsal) | no | no | yes |
 | Conflict detection (contradiction edges) | no | no | yes |
+| Multi-strategy RRF retrieval fusion | no | no | yes |
+| Cross-Encoder reranking | no | no | yes |
+| Explainable reasoning paths | no | no | yes |
+| Memory revision engine (sleep) | no | no | yes |
+| Hot/Warm/Cold memory tiering | no | no | yes |
+| Multimodal (text+image+audio) | no | no | yes |
 | Fuzzy recall ("I think I remember...") | no | no | yes |
 | Emergent abstraction (pattern discovery) | no | no | yes |
 | Temporal context (TimeSpine-indexed) | no | no | yes |
@@ -92,6 +103,9 @@ print(report)
 # Persist
 mgr.save("agent_memory.json")
 mgr.load("agent_memory.json")
+
+# Export to editable Markdown (GBrain-aligned)
+mgr.export_markdown("memories/")
 ```
 
 ## Architecture
@@ -109,12 +123,15 @@ Layer 3: Behavior    │  Cortex routing, memory gating, working memory,
                      │
 Layer 2: Cognitive   │  Hub abstraction, cascade recall, TimeSpine temporal index,
     (retriever.py,   │  sleep consolidation, evolution, ghost revival,
-     sleep.py,       │  abstraction, community detection, competition
-     evolution.py,   │  "How do memories connect, strengthen, and fade?"
-     ghost.py,       │
+     sleep.py,       │  abstraction, community detection, competition,
+     evolution.py,   │  conflict detection, memory revision, cross-encoder
+     ghost.py,       │  "How do memories connect, strengthen, and fade?"
      abstraction.py, │
      community.py,   │
      competition.py, │
+     conflict_detection.py,│
+     memory_revision.py,   │
+     cross_encoder.py,     │
      timespine.py,   │
      cascade.py,     │
      hub.py)         │
@@ -139,11 +156,14 @@ Layer 1: Storage     │  CRUD, persistence, ANN indexing, tiered storage,
 | `retrieval_pipeline.py` | 5-layer dimensional descent (L0→L4) with automatic degradation |
 | `scheduler.py` | Context-aware retrieval with memory type selection |
 | `working_memory.py` | Short-term buffer (15-item, 1h TTL) — auto-promotes to long-term |
-| `sleep.py` | 8-phase sleep: N1_Replay → N2_Merge → N3_Compression → N3b_AtomFacts → REM_Emotion → N4_Prune → N5_HubConnect → N6_IndexRebuild |
+| `sleep.py` | 10-phase sleep: N1_Replay → N2_Merge → N2b_ConflictDetection → N2c_MemoryRevision → N3_Compression → N3d_SleepRebuild → REM_Emotion → N4_Prune (v1.0.4) |
 | `evolution.py` | Survival-based decay (Ebbinghaus/Power-law/Exponential), belief transitions, interference |
 | `retriever.py` | HybridFusion + OscillationResonance + VectorSimilarity + Personalized PageRank + explainable scores |
 | `dual_channel.py` | System-1 (fast) + System-2 (deep) dual-channel retrieval with auto-trigger |
-| `bm25.py` | Sparse keyword retrieval (BM25) with reciprocal rank fusion for hybrid search |
+| `bm25.py` | Sparse keyword retrieval (BM25) with weighted reciprocal rank fusion for hybrid search |
+| `cross_encoder.py` | Post-retrieval Cross-Encoder reranking (sentence-transformers) for precision improvement |
+| `conflict_detection.py` | Semantic contradiction detection with overwrite/coexist/deprecate resolution strategies |
+| `memory_revision.py` | Sleep-phase memory revision engine — surprise-prioritized, template/LLM re-summarization |
 | `ghost.py` | Latent memory traces with fuzzy recall and contradiction tracking (NegativeGhost) |
 | `abstraction.py` | Emergent category discovery from anchor clusters |
 | `community.py` | Louvain community detection with centroid routing |
@@ -162,6 +182,11 @@ Layer 1: Storage     │  CRUD, persistence, ANN indexing, tiered storage,
 | `resonance.py` | Phase-locked oscillation resonance for temporal-coherent retrieval |
 | `streaming.py` | Streaming memory buffer with backpressure |
 | `benchmark.py` | Built-in benchmark suite (5 categories) |
+| `batch_vectorizer.py` | Deferred batch embedding writes (buffer >= 32 / >30s flush) with SQLite WAL crash recovery |
+| `zero_llm_pipeline.py` | 7-stage zero-LLM ingestion: security → embed → dedup → entity → classify → score → link |
+| `multimodal.py` | Cross-modal text/image/audio memory — CLIP image encoding, Whisper/spectrogram audio, unified embedding space |
+| `forget_certificate.py` | Ed25519-signed JWS deletion certificates — GDPR Article 17 provable erasure |
+| `markdown_export.py` | Operator-owned plain-text memory export (GBrain-aligned); timeline + topic Markdown |
 | `config.py` | Centralized YAML config with schema validation, dot-path access, overrides |
 
 ## Retrieval Benchmarks
@@ -264,11 +289,14 @@ See `star_graph/defaults.yaml` for all 300+ tunable parameters.
 # From PyPI (package name: NWcortex, import as: star_graph)
 pip install NWcortex
 
-# With sentence-transformers for semantic embeddings (~2GB model download on first use)
+# With sentence-transformers for semantic embeddings
 pip install "NWcortex[embeddings]"
 
-# Optional: for SQLite storage backend
-pip install aiosqlite
+# With MCP server support (for AI agent integration)
+pip install "NWcortex[mcp]"
+
+# Everything
+pip install "NWcortex[all]"
 
 # Run demo
 python examples/emergence_demo.py
@@ -276,6 +304,91 @@ python examples/emergence_demo.py
 
 **Note:** Without `[embeddings]`, the system uses a lightweight TF-IDF fallback for text encoding. Install `sentence-transformers` only if you need semantic-quality embeddings.
 
+## MCP Server — AI Agent Integration
+
+NeuroWeave Cortex ships with a [Model Context Protocol](https://modelcontextprotocol.io) server. Connect any MCP-compatible AI agent (Claude Desktop, OpenClaw, Cursor, etc.) for persistent cognitive memory across conversations.
+
+### Quick start
+
+```bash
+# Install with MCP support
+pip install "NWcortex[mcp]"
+
+# Launch the MCP server on stdio
+nwc-mcp
+
+# With persistent storage
+nwc-mcp --storage agent_memory.json
+
+# Load a previously saved memory graph
+nwc-mcp --load my_memories.json
+```
+
+### Tool reference (12 tools)
+
+| Tool | Description |
+|------|-------------|
+| `remember` | Store a long-term memory (tags, importance, emotional valence) |
+| `remember_working` | Store in fast ephemeral working-memory buffer (current task context) |
+| `recall` | Context-aware semantic retrieval with task-type routing |
+| `forget` | Remove a memory, creating a ghost trace for potential fuzzy recall |
+| `sleep` | Run 5-phase sleep consolidation — merges, prunes, forms schemas |
+| `consolidate` | Micro-consolidation — incremental, non-blocking |
+| `stats` | Memory system statistics (anchors, edges, ghosts, schemas, cognitive health) |
+| `fuzzy_recall` | Low-confidence recall from ghost traces ("I seem to remember...") |
+| `get_profile` | Inferred user profile from accumulated memories |
+| `evolve` | Memory evolution cycle (decay, boost, conflict resolution) |
+| `save` | Persist memory graph to disk (JSON) |
+| `load` | Load memory graph from disk (JSON) |
+
+### Claude Desktop config
+
+```json
+{
+  "mcpServers": {
+    "neuroweave-cortex": {
+      "command": "nwc-mcp",
+      "args": ["--storage", "/path/to/agent_memory.json"]
+    }
+  }
+}
+```
+
+### OpenClaw config
+
+```json
+{
+  "mcpServers": {
+    "neuroweave-cortex": {
+      "command": "nwc-mcp",
+      "args": ["--storage", "/path/to/agent_memory.json"]
+    }
+  }
+}
+```
+
+## REST API Server
+
+```bash
+# Start the REST server
+nwc-server --port 8420
+
+# Or via module
+python -m star_graph.server --port 8420
+```
+
+Endpoints:
+
+| Method | Path | Description |
+|--------|------|-------------|
+| GET | `/health` | Health check with anchor/edge counts |
+| GET | `/metrics` | Prometheus-format metrics |
+| GET | `/stats` | Full memory system statistics |
+| POST | `/remember` | Store a memory `{"text": "...", "tags": [...]}` |
+| POST | `/recall` | Retrieve memories `{"query": "...", "max_items": 10}` |
+| POST | `/sleep` | Run sleep consolidation |
+| POST | `/consolidate` | Run micro-consolidation |
+
 ## Interactive Demo
 
 ```bash
@@ -287,6 +400,9 @@ python examples/emergence_demo.py
 # sg-query --trace "When did Alice visit Hawaii?"
 # sg-stats --schemas --ghosts
 # sg-sleep --retention 0.15 --edge-prune 0.1
+# nwc-export --output memories/ --organize both
+# nwc-forget <anchor_id> --certificate --reason gdpr_art17
+# nwc-verify certificates/forget-<id>.jws
 ```
 
 ## Benchmarks
@@ -306,7 +422,7 @@ pytest tests/ -v
 pytest tests/ --cov=star_graph --cov-report=term
 ```
 
-**Status:** 1,989 tests passing, 80% coverage (v1.0.0).
+**Status:** 1,989 tests passing, 84 modules (v1.0.4-dev).
 
 ## Roadmap