WARM
Memory serverOfficial Writerslogic project
Created by Writerslogic
Published in writerslogic/warm. Publisher page
27 GitHub stars
What it is
WritersLogic Adaptive Retention & Memory (WARM): local document search (BM25 + embedding cosine) and vector-symbolic associative memory.
AutoGineer overview, from the brain’s own files
WARM (WritersLogic Adaptive Retention & Memory) is a local-first memory engine designed to provide AI agents and applications with persistent storage and retrieval. It combines traditional document search—using stemmed BM25 and vector-based cosine similarity—with a vector-symbolic layer for associative memory and multi-hop queries. The system is built in Rust and can be deployed as an embedded library, an HTTP server, or an MCP server, allowing it to integrate with various agent hosts and development environments. It emphasizes local control, offering optional encryption and provenance features while avoiding external service calls or telemetry.
How it works
Memory model: Hybrid
AutoGineer overview, from the brain’s own files
The system stores text, vectors, and facts in a local directory. It uses a hybrid approach where document search combines chunking and BM25 with vector embeddings. The vector-symbolic layer utilizes sparse hypervectors to support associative memory, triples, and multi-hop queries. It provides an API for memory management including remember, recall, and forget operations, with support for source provenance and atomic storage updates.
| Folder | Role | Files |
|---|---|---|
| .github | agent-config | 37 |
| assets | attachments | 10 |
| docs | docs | 1850 |
| examples | examples | 47 |
| scripts | scripts | 77 |
| src | source-code | 327 |
| tests | tests | 362 |
| benches | folder | 7 |
| benchmarks | folder | 2898 |
| bindings | folder | 273 |
| deploy | folder | 18 |
| figures | folder | 6 |
| fuzz | folder | 54 |
| integrations | folder | 225 |
| interop | folder | 3 |
| notebooks | folder | 3 |
| npm | folder | 1 |
| proptest-regressions | folder | 1 |
| schemas | folder | 9 |
The agent reads
- query_batch evidence: README.md
- get_document evidence: README.md
- list_documents evidence: README.md
- get_document_info evidence: README.md
- list_document_info evidence: README.md
Requirements
Runtimes: Node.js, Python >=3.9, Rust
Databases: SQLite
API keys (by name): WARM_API_KEY
Works with: Gemini CLI
Set it up
This brain runs code automatically: 1 hook (claude-code), 1 script. Review benchmarks/results/sbg_positions_development/setup.sh, integrations/editors/claude-code/plugin/hooks/hooks.json before opening the folder in Claude Code. No step below turns a hook on for you.
benchmarks/results/sbg_positions_development/setup.sh:sh benchmarks/results/sbg_positions_development/setup.shintegrations/editors/claude-code/plugin/hooks/hooks.json:node
- Not scanned by AutoGineerThe author’s README also describes installing it from a clone of its repository. Read its install section: AutoGineer did not scan that path. https://github.com/writerslogic/warm/blob/e176d5572dd86bf2233bcb120e3a485dc109ea3a/README.md#get-started
- Not scanned by AutoGineerThe author’s README also describes installing it as the npm package writerslogic-warm. Read its install section: AutoGineer did not scan that path. https://github.com/writerslogic/warm/blob/e176d5572dd86bf2233bcb120e3a485dc109ea3a/README.md#get-started
Created by Writerslogic. Published in writerslogic/warm. Licence: Apache License 2.0.
Usage
AutoGineer overview, from the brain’s own files
- Integrating memory into an AI agent via the MCP server.
- Building local-first applications that require embedded search and storage.
- Providing a multi-tenant memory service via the HTTP server.
- Using the Rust library for in-process memory management.
- “Remember that the project deadline is Friday.”
- The agent stores the fact with associated metadata, making it available for future retrieval.
- “Recall information about the project deadline.”
- The agent retrieves the stored fact using hybrid search and provides the information with source provenance.
Documentation by Writerslogic
<img src="https://raw.githubusercontent.com/writerslogic/warm/main/assets/crystal-halo/crystal-detailed-static.png" alt="WARM logo" width="96" align="left">
WARM
Local memory for AI agents and applications.
WARM (WritersLogic Adaptive Retention & Memory) gives agents persistent memory, full-text and vector search, and source provenance in one embeddable Rust engine. Use it inside your application, through HTTP, or from an MCP client, with storage on infrastructure you control.
Image: CI Image: MSRV 1.89 Image: OpenSSF Scorecard Image: OpenSSF Best Practices Image: License: Apache-2.0 Image: WARM MCP server: quality and maintenance score on Glama
Website · Docs · Get started · Standards and provenance · Evaluation · Capabilities · Limits · Contribute
<br clear="left">
What it is
WARM is a Rust library that stores text, vectors and facts in a directory on your machine and finds them again. It ships with optional Node.js, C, WASM and UniFFI bindings, an HTTP server with portable JavaScript and Python clients, and an MCP server. The embedded engine makes no service calls or telemetry requests. HTTP clients connect to the server you configure; model stages are opt-in.
| Capability | What you get |
|---|---|
| Document search | Chunking, stemmed BM25 over blocked postings, Boolean and phrase queries, hybrid fusion with embeddings you supply, metadata filters, byte offsets |
| Vector index | Graph index with a memory-mapped vertex file, an updatable delta for inserts and deletes, optional query-aware routing trained on sample queries |
| Memory API | remember, recall, forget and explain with time validity and supersession; conversation memory with source provenance |
| Storage | Atomic replace and delete, crash-safe journal and checkpoints, verified compaction, one writer per store |
| Security | Optional AES-256-GCM storage encryption with documented coverage and limits, signed audit log, COSE/SCITT provenance |
| Vector-symbolic layer | Sparse hypervectors with bind, bundle and permute, triples with multi-hop queries, Hopfield cleanup |
Who it's for
| You are building | What you get |
|---|---|
| An agent or assistant that needs memory | A memory API with forgetting that removes facts from the store, and an MCP server for editors and agent hosts |
| A local-first or offline app | Embedded search and storage in one directory, with no server process and optional encryption |
| A service that needs vector or text search | warm-server over HTTP with tenants and scoped keys, or the library in-process |
Get started
For an editor or agent host, start with the MCP integration. For an application, choose the Rust library or HTTP server below. WARM runs locally or on your own infrastructure; this repository does not provide a managed cloud service.
To try the offline memory demo from source:
git clone https://github.com/writerslogic/warm
cd warm
cargo run --release --example demo-offline-agent<details open> <summary><strong>Rust</strong></summary>
From your own Rust project, add the current Apache-2.0 source. The crate is warm-rust; the library is imported as writerslogic_warm. The MSRV is Rust 1.89; local-models and --all-features need Rust 1.94 or newer.
cargo add warm-rustThe historical v0.6.1 source
remains AGPL-3.0-or-later; it is available from that Git tag, not as a
warm-rust 0.6.1 release on crates.io.
use writerslogic_warm::{DocumentInput, WarmCore, SearchOptions};
let dir = tempfile::tempdir()?;
let warm = WarmCore::new(4096, Some(dir.path().display().to_string()), None)?;
warm.memorize_document(DocumentInput {
id: "backups".into(),
text: "Restore deleted documents from verified backups.".into(),
source_uri: Some("operations.md".into()),
version: Some("1".into()),
metadata: Some(serde_json::json!({ "project": "alpha" })),
chunk_words: None, overlap_words: None, store_text: None, embeddings: None, title_weight: None,
})?;
let hits = warm.search_documents("restore backups", &SearchOptions::default())?;
println!("{} {:?}", hits[0].document_id, hits[0].source_uri);Rust API change in 0.6.6 (0.6.5 returns these values directly): WarmCore::query, query_batch, and
analyze_components return Result. Handle the result with ? or an explicit
error branch; deferred-encoding failures no longer become successful partial
query results. get_document, list_documents, get_document_info, and list_document_info
also return Result; retained-text and checkpoint decoding failures propagate
instead of panicking, and reads no longer hydrate the
entire checkpoint or hold store locks during text I/O.
</details>
<details> <summary><strong>JavaScript and TypeScript: portable client</strong></summary>
warm-sdk 0.2.1 is the portable HTTP client for a separately running warm-server.
It has no native binaries or runtime dependencies and uses standard Fetch and Web Streams APIs
(Node 18 or newer). Installing or importing it does not load the embedded engine or download models.
For a published release, install the client in your application:
npm install [email protected]Check the available npm versions. To build the client from this checkout:
npm ci --prefix bindings/js/warm-sdk
npm run build --prefix bindings/js/warm-sdkIn your application, install that directory with
npm install /absolute/path/to/warm/bindings/js/warm-sdk.
Start the server in a separate terminal:
cargo build --release --features server --bin warm-server
WARM_API_KEY=change-me target/release/warm-server --input-dim 384import { WarmClient } from 'warm-sdk'
const warm = new WarmClient({
url: process.env.WARM_URL ?? 'http://127.0.0.1:8080',
apiKey: process.env.WARM_API_KEY, tenantId: 'demo',
})
await warm.memory.remember('Restore deleted documents from verified backups.', {
id: 'backups', metadata: { project: 'alpha' },
})
const hits = await warm.memory.recall('restore backups', { filter: { project: 'alpha' } })
console.log(hits[0]?.text)Set WARM_API_KEY to the server's key before running the client. This memory example uses
lexical retrieval and needs no embedding model. A runnable source example also checks retrieval
and deletion in an isolated namespace:
WARM_URL=http://127.0.0.1:8080 WARM_API_KEY=change-me \
node examples/quickstart/js-client/quickstart.mjsSee the client API and release availability. Browser deployments must permit the application's origin at the HTTP service boundary.
</details>
<details> <summary><strong>Node.js: optional embedded engine</strong></summary>
writerslogic-warm embeds the Rust engine in Node. For a published engine release,
install it explicitly with npm's optional dependencies enabled:
npm install [email protected]Check the available npm versions. To build from the cloned repository, with Rust installed:
npm ci --omit=optional
npm run buildThe portable client and embedded engine are separate packages with different APIs.
listDocuments supports scoped pruning. See the
integration guide for source
provenance, namespace isolation, dry runs and deletion limits.
const { WarmMemorySystem } = require('writerslogic-warm')
const { DocumentMemory } = require('writerslogic-warm/semantic')
async function main() {
const dir = require('node:fs').mkdtempSync(require('node:path').join(require('node:os').tmpdir(), 'warm-readme-'))
const memory = new DocumentMemory(new WarmMemorySystem(4096, dir))
await memory.memorize({
id: 'backups', text: 'Restore deleted documents from verified backups.',
sourceUri: 'operations.md', version: '1', metadata: { project: 'alpha' },
})
const [hit] = await memory.search('restore backups', { filter: { project: 'alpha' } })
console.log(hit.documentId, hit.sourceUri, hit.text)
await memory.flush()
}
main().catch(error => { console.error(error); process.exit(1) })</details>
<details> <summary><strong>Server and Python</strong></summary>
The Python package warm-sdk is an HTTP client for warm-server, which stores vectors you
supply. (warm-python, imported as writerslogic_warm, provides native vector indexes and phase vectors, but no document search.)
cargo build --release --features server --bin warm-server
WARM_API_KEY=change-me target/release/warm-server --input-dim 384 &
uv pip install warm-sdkfrom writerslogic_warm_sdk import WarmClient
client = WarmClient(url="http://127.0.0.1:8080", api_key="change-me", tenant_id="demo")
client.add_documents([{"id": "a", "vector": [0.1] * 384, "text": "hello", "metadata": {}}])
print(client.query([0.1] * 384, top_k=1)["matches"][0]["id"])Endpoints, flags, tenants and limits are in the
server docs. A runnable Python quickstart
with a toy embedder is in examples/quickstart/python.
</details>
<details> <summary><strong>MCP</strong></summary>
warm-mcp serves a store to any Model Context Protocol client over stdio or HTTP:
npx -y writerslogic-warm-mcp # stdio; downloads a verified binary on first run
npx -y writerslogic-warm-mcp --transport http # http://127.0.0.1:8765/mcpThe MCP package is a separate launcher. To build the server from source:
cargo build --release --features mcp,security --bin warm-mcp # add local-models for semantic recall
./target/release/warm-mcp # stdio
./target/release/warm-mcp --transport http # http://127.0.0.1:8765/mcpClient setup, data directory and the one-writer rule: MCP integration.
For normalized content from multiple applications, see the organizational knowledge adapter.
</details>
The five-minute quickstart
walks through ingest, search, a SQLite FTS5 comparison and encryption in Rust, Python and Node,
and CI runs that code on every pull request. For a whole agent memory with no embeddings and no
network, run cargo run --example demo-offline-agent.
<a id="headline-results"></a> <a id="where-its-behind"></a>
Evaluation
Recorded results and limitations cover memory quality, text retrieval and vector search. Each result retains its workload, hardware, model, source revision and known losses. The full comparison and reproduction protocol provide the evidence for evaluating WARM on your own workload.
What ships today
| Capability | Status | Documented limit |
|---|---|---|
| Hybrid document search (BM25 + exact cosine, filters, byte offsets) | Ships | Exact scan of eligible chunks; benchmark your corpus before assuming capacity (guide) |
Boolean lexical search (+required, -excluded, quoted phrases) | Ships | Flat queries only, no nesting; an empty or only-excluded query matches nothing (design) |
| Atomic replace/delete, crash-safe journal, verified compaction, process lock | Ships | One writer per store; compaction is not secure erasure |
Vector index (qgraph vertex graph, memory-mapped) | Ships | Build time and serving RSS behind hnswlib on SIFT-1M (gaps) |
| Updatable vector index (durable delta for inserts and deletes) | Ships | Single writer; format in VECTOR-UPDATES.md |
| Query-aware graph routing (trained on sample queries) | Ships (opt-in) | Needs representative training queries; tested on VIBE llama and yi (report) |
| Memory API: remember, recall, forget, explain | Ships | Forgetting removes a fact from the store; it does not cover what a model inferred (docs) |
Conversation memory (conversation-memory) | Ships | Bounded context with exact source provenance; native recall not yet benchmarked on LongMemEval end to end |
| Meaning memory: triples, multi-hop, Hopfield cleanup | Ships (runtime opt-in) | Tested on synthetic and 2,120-fact sets (research_bench_16384_256.json) |
Union-bundle membership (experimental) | Research | At D = 16,384: positive gap to N = 500 (4% false positives), tested to 2,500 (benchmark_scaling_results.json) |
Sparse vector index (memorize_vector) | Experimental | Not competitive with HNSW (results) |
Encrypted stores (security) | Ships | AES-256-GCM, Argon2id; covers the log, index caches and the text segment. Metadata, file names and sizes are not encrypted |
Signed audit log, credential-gated access, COSE/SCITT/C2PA (provenance) | Ships | Host must authenticate DIDs; PoSME, RATS, did:web not implemented |
On-device models (local-models: Qwen3 embed/rerank/LLM) | Ships | Local weights only; never downloads |
warm-server HTTP API, multi-tenant (server) | Ships | Single node; TLS, per-tenant scoped keys (guide) |
MCP server (mcp: warm-mcp, stdio and HTTP) | Ships | Encryption needs security; semantic recall needs local-models |
JavaScript/TypeScript client (warm-sdk 0.2.1) | Source and release builds | Portable HTTP client; requires a separately running server |
Embedded Node.js (writerslogic-warm 0.6.8) | Source and release builds | Eight optional native platform packages; separate from the client |
| C, Go, .NET, Java, Kotlin, Swift and WebAssembly bindings | Source available | Require local builds; the language packages and native bundles are not yet published (runtime guides) |
| Python | HTTP client (warm-sdk); native vector indexes and phase vectors (warm-python) | No in-process document search |
Query-private search (private-search) | Experimental | For searching an index someone else holds; hides the query vector from the scorer only (notes) |
Optional engine capabilities use Cargo features or runtime configuration; Cargo features lists the build choices.
The compatibility policy defines API stability, deprecation windows and storage-format migration rules.
Architecture
flowchart TB
subgraph bindings["Bindings and servers"]
rust["Rust crate"]
node["Node: writerslogic-warm (napi)"]
ffi["C, WASM, UniFFI"]
clients["JavaScript and Python: warm-sdk"]
server["warm-server (HTTP)"]
mcp["warm-mcp (stdio, HTTP)"]
end
subgraph engine["Engine API"]
core["WarmCore: documents, vectors, facts, mutation gate"]
prov["Security and provenance: signed audit log, COSE/SCITT"]
end
subgraph indexes["Indexes and memory"]
text["Text index: blocked postings, BM25, Boolean and phrase"]
vector["Vector index (qgraph): vertex file, updatable delta, query-aware projection"]
memory["Memory API: remember, recall, forget, conversation memory"]
vsa["Vector-symbolic layer: bind, bundle, Hopfield cleanup"]
end
subgraph storage["Storage"]
journal["Journal and checkpoints"]
segments["Append-only segments (CRC32, LZ4)"]
mapped["Memory-mapped files"]
crypto["AES-256-GCM encryption (optional)"]
end
clients --> server
rust --> core
node --> core
ffi --> core
server --> core
mcp --> core
core --> text
core --> vector
core --> memory
core --> vsa
core --> prov
memory --> text
text --> journal
vector --> journal
memory --> journal
journal --> segments
segments --> mapped
segments --> cryptoModule map: docs/ARCHITECTURE.md.
Limits
Not anonymous. Hypervector sketches and term counts are not encryption; anyone with a store can encode candidate text and compare.
storeText: falsekeeps term counts, IDs and embeddings, so it is not a privacy control.Document search is conventional. It does not use the hypervectors; the holographic parts are the algebra and the meaning memory.
One writer per store. A second writing process fails at open with a lock error.
Capacity numbers are points, not guarantees, and are stated only over the tested range (capacity).
Benchmark results are single machines. Conditions are in each results file; the single-core ann-benchmarks run and the VIBE high-recall points are losses (see Where it's behind).
Masking is obfuscation. Client-side vector masking preserves every pairwise similarity.
A store is a directory and
WarmCore::new(.., None, ..)uses the current directory.
Full statement: docs/PRIVACY.md. Report vulnerabilities privately via SECURITY.md.
Contributing
CONTRIBUTING.md: setup, the checks CI runs, DCO sign-off (git commit -s, no
CLA). Start with a good first issue
or Discussions. See the
Code of Conduct. Benchmark changes regenerate the results file they touch;
the README cites files, not remembered figures.
License
Apache-2.0. See LICENSE. Use, modify and embed it in open-source or proprietary software, including network services, with no copyleft obligation. The historical v0.6.1 source remains available under its AGPL-3.0-or-later license.
docs/ARCHITECTURE.md
Architecture
Module Graph
lib.rs Public Rust exports and module declarations
|
bindings/
+-- node/ Node.js bindings, configuration, DTOs and local embedder
+-- python/ PyO3 entry point, vector indexes and phase-vector bindings
|
tests/ Engine, security, meaning and export tests
|
bin/ Shipped command-line tools and servers
+-- research/ Benchmark, diagnostic and experiment executables
|
core/
+-- engine/
| +-- mod.rs WarmCore: main orchestrator
| +-- mutation.rs Validated, logged transactions and replay
| +-- document_api.rs Document lifecycle and bounded batches
| +-- query.rs Query routing and execution
| +-- router.rs Adaptive retrieval strategy selection
| +-- shard.rs ShardSet, ShardManager, Shard
| +-- concepts.rs Concept synthesis (clustering + bundling)
| +-- knowledge.rs Triplets, sequences, analogies
| +-- structural.rs Fuzzy structural queries (algebraic + materialized)
| +-- multi_hop.rs Multi-hop reasoning (rule rewrite + chained lookup)
|
+-- cognition/
| +-- mod.rs CognitionLoop: background discovery engine
| +-- patterns.rs PatternScanner: relation co-occurrence analysis
| +-- abstraction.rs AbstractionEngine: prototype concept discovery
| +-- gaps.rs GapDetector: epistemic gap detection
| +-- hypothesis.rs HypothesisEngine: gap-filler proposals
| +-- analogy.rs AnalogyDetector: structural isomorphism
| +-- governor.rs MemoryGovernor: dedup, forgetting, IDF refresh
| +-- refiner.rs DistributionalRefiner: self-organizing atom vectors
| +-- loop.rs Background thread lifecycle
|
+-- agency/
| +-- mod.rs Goal-directed reasoning layer
| +-- goals.rs Goal definition and lifecycle
| +-- planner.rs Plan generation from goals
| +-- questions.rs Question generation for knowledge gaps
| +-- self_modify.rs Self-modification proposals
|
+-- entangled.rs EntangledHVec: sparse binary hypervector type
+-- encoding.rs Text -> sparse vector (multiscale words and character n-grams)
+-- experimental/ Opt-in codebooks, block codes, Bloom memory and phase graphs
+-- hopfield.rs Modern Hopfield network with sparse softmax
+-- phase_resonator.rs Resonator network for symbolic factorization
+-- decompose.rs Decomposer: vector decomposition
+-- graph.rs Graph engine: typed relations, multi-hop BFS, temporal
+-- documents.rs Chunking, stemmed BM25, metadata filters, BM25/cosine score blend
+-- schema.rs Store and embedding-space compatibility
+-- durable_file.rs Flushed atomic file publication
+-- storage.rs PersistentArena: mmap segmented log and generations
+-- config.rs WarmConfig, MeaningConfig, CognitionConfig, and sub-configs
+-- security.rs SigningManager, EncryptionManager (feature-gated)
+-- audit.rs AuditLog: append-only operation log
+-- diffusion.rs DiffusionFactorizer: score-based vector decomposition
+-- text.rs TextProcessor: readability metrics
+-- types.rs Shared types (RetrievalResult, ConceptCandidate, etc.)
+-- error.rs WarmError enum
+-- intersection.rs Sparse sorted-merge intersection
+-- atom_memory.rs AtomMemory: concept vector store (meaning memory)
+-- composite_memory.rs CompositeMemory: role-bound composite vectors
+-- triple_store.rs TripleStore: symbolic (S, R, O) index
+-- role.rs RoleRegistry: role-shift algebra
+-- rules.rs RuleStore: composition rule definitions
+-- admission.rs AdmissionControl: fan-out gating
+-- indexed_memory.rs IndexedMemory: posting + IDF substrate
+-- posting.rs PostingShard: per-dimension posting lists
+-- idf.rs IdfWeights: IDF with proportional clipping
+-- tombstone.rs TombstoneMap: soft-delete tracking
+-- index/
| +-- mod.rs Index traits
| +-- inverted.rs Sparse inverted index for high-sparsity queries
+-- ivf/
| +-- mod.rs IVFIndex: inverted file with product quantization
| +-- training.rs IVF training pipeline
| +-- query.rs IVF query execution
| +-- kmeans.rs K-means clustering
| +-- pq.rs Product quantization
| +-- nystrom.rs Nystrom dimensionality reduction
| +-- inverted_list.rs Inverted list storage
+-- nsg/
+-- mod.rs NSGIndex: navigable small-world graph
+-- training.rs NSG construction
+-- search.rs Greedy graph search
+-- graph.rs Graph operations (KNN, pruning, centroid)The crate root keeps the existing public Rust exports. Experimental modules live under
core/experimental/; their existing core::module_name imports remain available through
re-exports. Cargo keeps every binary name and required feature unchanged. Historical
benchmark manifests retain the paths at their measured revisions.
Meaning Memory Module Graph
WarmCore (engine/mod.rs)
|
+-- atom_memory.rs AtomMemory: concept vector store
| +-- indexed_memory.rs IndexedMemory: posting lists + IDF + tombstones
| +-- posting.rs PostingShard: inverted posting lists per dimension
| +-- idf.rs IdfWeights: IDF weighting with proportional clipping
| +-- tombstone.rs TombstoneMap: soft-delete tracking
|
+-- composite_memory.rs CompositeMemory: role-bound triple vectors
| +-- indexed_memory.rs (shared substrate with AtomMemory)
|
+-- triple_store.rs TripleStore: symbolic (S, R, O) index
| Four-way FxHash index: by_subject, by_relation,
| by_object, by_composite
|
+-- role.rs RoleRegistry: role -> cyclic-shift mapping
| compose(), unbind(), compose_triple()
|
+-- rules.rs RuleStore: CompositionRule definitions
| Maps relation chains to derived relations
|
+-- admission.rs AdmissionControl: fan-out gating
| Algebraic vs. MaterializedLookup decision
|
+-- decompose.rs Decomposer: vector decomposition
|
+-- engine/
+-- structural.rs fuzzy_structural_query(): algebraic + materialized paths
+-- multi_hop.rs multi_hop_query(): rule rewrite + chained lookupData Flow: Structural Query
fuzzy_structural_query(known_bindings, target_role)
-> RoleRegistry.compose(known) // build partial query vector
-> CompositeMemory.overlap_scan(query) // IDF-weighted posting intersection
-> AdmissionControl.check(fan_out) // gate on candidate count
|
+-- Algebraic path (fan_out <= limit):
| -> composite.bind(query) // XOR-unbind known roles
| -> permute(dim - target_shift) // inverse cyclic shift
| -> hopfield_cleanup(residual) // attractor network recovery
| -> overlap_scan + sparse_softmax + iterate
| -> return (entity_id, confidence)
|
+-- Materialized path (fan_out > limit):
-> TripleStore.by_composite_id() // symbolic index lookup
-> extract target_role field
-> return (entity_id, score)Data Flow: Multi-Hop Query
multi_hop_query(start, [rel1, rel2, ...], ctx, rule_store)
-> if single relation:
single_hop -> fuzzy_structural_query
-> if two relations + matching CompositionRule:
rule_rewrite -> fuzzy_structural_query(derived_relation)
-> else:
chained_lookup -> TripleStore walk, hop by hopTransactions and lock ordering
Persistent vector, document, graph, triple, and rule mutations take mutation_gate.write()
before the shard/component locks. Main vector/document/structural/graph queries take its read
lock. A mutation validates its complete operation list, logs one framed transaction, then
applies every operation to memory. This gate also serializes compaction and index publication.
Audit and provenance sidecars are separate operations after the main transaction.
Within a shard, vector/registry/posting changes and cache invalidation follow the shard's lock order. Index training copies a consistent snapshot, releases the gate while training, then checks the revision before publishing. A stale training run returns an explicit retry error.
Write and recovery paths
public mutation
-> validate dimensions, identifiers, limits, and complete operation list
-> acquire transaction write gate and shard write lock
-> serialize HMS-TXN-2 + typed mutations as one record
-> optional authenticated encryption
-> CRC-framed append and mmap flush
-> apply vectors, postings, documents, triples, relations, and rule changes
-> increment data revisionRecovery validates the schema before opening data, resolves the active CURRENT generation,
validates arena frames, decrypts authenticated payloads when configured, and replays typed
transactions in order. Main-data decryption/replay failures are errors. ANN caches are optional:
only checksummed files with matching shard topology and arena checkpoints are loaded. Invalid
or stale caches are discarded; inverted indices are reconstructed from live vectors.
Compaction
Compaction excludes persistent mutations and main queries. It snapshots live vectors, documents,
atoms, composites, triples, relation types, relations, rules, and derived-fact ownership into a
new arena. It flushes and verifies the arena, installs a generation directory, then publishes
CURRENT with an atomic file replacement. Old files are cleaned up only after publication.
A process interruption therefore selects either the complete old or complete new generation.
Directory publication is synced on Unix; platform/filesystem durability limits are documented
in production-readiness.md.
Document and meaning-derived chunk IDs have reserved prefixes. Updating or deleting a document removes its owned chunks. Replacing or deleting a meaning source removes its owned inferred triples; the ownership mapping survives compaction and recovery. Explicit triplet replacement removes the prior structural record before inserting the new one.
Retrieval paths
Sparse vector queries use the adaptive router to select brute-force, inverted, NSG, or IVF
retrieval. Replacements and deletions update postings incrementally, reuse slots, and invalidate
incompatible ANN caches. No training is hidden inside ingestion. indexStatus reports
recommendations; the application schedules maintainIndices.
Document queries first filter metadata/source/document IDs, then compute BM25 over lexical terms and optional exact cosine over normalized stored dense embeddings. Weighted reciprocal rank fusion combines the rankings. The optional JavaScript adapter embeds queries locally and reranks candidate passages with a local ONNX cross-encoder. This document path currently scans eligible chunks; it does not use the sparse ANN graph as a dense-vector approximation.
semantic.js batches model inference and provides a bounded operation queue. Async iterable
ingestion awaits each document, preventing unbounded pending native/model work. Model identity,
artifact revision, preprocessing fingerprint, dimensions, normalization, and metric must match
the store's embedding space.
Validation
tests/reliability.rs covers signed projections, structural answers, durable deletes, ANN
restart/corruption handling, and concurrent writes/compaction. tests/documents.rs covers source
offsets, filters, versions, semantic candidates, bounded ingestion, and atomic source migration.
Storage unit tests inject child-process exits around compaction publication. Native Node tests
exercise actual CJS/ESM bindings, encryption failure, file ingestion, and queue backpressure.
A packed-install smoke test loads the distributable native module and document adapter.
512 files · 481 not served here
- .bestpractices.json (not served here)
- .dockerignore (not served here)
- .editorconfig (not served here)
- .gitattributes (not served here)
- .github/CODEOWNERS (not served here)
- .github/FUNDING.yml (not served here)
- .github/ISSUE_TEMPLATE/bug_report.yml (not served here)
- .github/ISSUE_TEMPLATE/config.yml (not served here)
- .github/ISSUE_TEMPLATE/feature_request.yml (not served here)
- .github/SECURITY.md (not served here)
- .github/SUPPORT.md (not served here)
- .github/badges/coverage.json (not served here)
- .github/dependabot.yml (not served here)
- .github/pull_request_template.md (not served here)
- .github/workflows/bench-refresh.yml (not served here)
- .github/workflows/bindings.yml (not served here)
- .github/workflows/changelog.yml (not served here)
- .github/workflows/ci.yml (not served here)
- .github/workflows/coverage.yml (not served here)
- .github/workflows/dco.yml (not served here)
- .github/workflows/dependency-review.yml (not served here)
- .github/workflows/docs.yml (not served here)
- .github/workflows/examples.yml (not served here)
- .github/workflows/format.yml (not served here)
- .github/workflows/fuzz.yml (not served here)
- .github/workflows/importers.yml (not served here)
- .github/workflows/js-sdk.yml (not served here)
- .github/workflows/links.yml (not served here)
- .github/workflows/mcp-container.yml (not served here)
- .github/workflows/npm-packages.yml (not served here)
- .github/workflows/python-wheels.yml (not served here)
- .github/workflows/release.yml (not served here)
- .github/workflows/reproducible.yml (not served here)
- .github/workflows/retire-legacy-npm.yml (not served here)
- .github/workflows/scorecard.yml (not served here)
- .github/workflows/standards-interop.yml (not served here)
- .github/workflows/warm-js-sdk-release.yml (not served here)
- .github/workflows/warm-mcp-release.yml (not served here)
- .github/workflows/warm-models-release.yml (not served here)
- .github/workflows/warm-py-release.yml (not served here)
- .github/workflows/warm-sdk-release.yml (not served here)
- .gitignore (not served here)
- .lychee.toml (not served here)
- .npmignore (not served here)
- ATTRIBUTION.md
- AUTOGINEER-RUNS.md
- CHANGELOG.md
- CITATION.cff (not served here)
- CODE_OF_CONDUCT.md
- CONTRIBUTING.md
- Cargo.lock (not served here)
- Cargo.toml (not served here)
- Dockerfile (not served here)
- LICENSE
- README.md
- SECURITY.md
- assets/crystal-halo/README.txt (not served here)
- assets/crystal-halo/crystal-detailed-loop.apng (not served here)
- assets/crystal-halo/crystal-detailed-loop.svg (not served here)
- assets/crystal-halo/crystal-detailed-loop.webp (not served here)
- assets/crystal-halo/crystal-detailed-static.png (not served here)
- assets/crystal-halo/crystal-detailed-static.svg (not served here)
- assets/crystal-halo/source/build_additional_exports.py (not served here)
- assets/crystal-halo/source/build_crystal_halo.py (not served here)
- assets/crystal-halo/source/crystal-halo-geometry.json (not served here)
- assets/social-preview.png (not served here)
- benches/audit_delivery.rs (not served here)
- benches/ingest_alloc.rs (not served here)
- benches/intersection_kernels.rs (not served here)
- benches/query_alloc.rs (not served here)
- benches/vector_record_query.rs (not served here)
- benches/vector_updates.rs (not served here)
- benches/warm_benchmark.rs (not served here)
- benchmarks/ann-benchmarks/campaign/ga.py (not served here)
- benchmarks/ann-benchmarks/campaign/gen_heldout.py (not served here)
- benchmarks/ann-benchmarks/campaign/mkdefs.py (not served here)
- benchmarks/ann-benchmarks/campaign/mkdefs_ivf.py (not served here)
- benchmarks/ann-benchmarks/campaign/mkreport.py (not served here)
- benchmarks/ann-benchmarks/campaign/plans/b1-sift.json (not served here)
- benchmarks/ann-benchmarks/campaign/plans/b2-glove.json (not served here)
- benchmarks/ann-benchmarks/campaign/plans/g1-sift.json (not served here)
- benchmarks/ann-benchmarks/campaign/prebuild.py (not served here)
- benchmarks/ann-benchmarks/campaign/prof.sh (not served here)
- benchmarks/ann-benchmarks/campaign/profq.py (not served here)
- benchmarks/ann-benchmarks/campaign/run_gen.py (not served here)
- benchmarks/ann-benchmarks/campaign/summ.py (not served here)
- benchmarks/ann-benchmarks/campaign/test_campaign_integrity.py (not served here)
- benchmarks/ann-benchmarks/campaign/warm_module_campaign.py (not served here)
- benchmarks/ann-benchmarks/campaign/warmivf_module.py (not served here)
- benchmarks/ann-benchmarks/warm/Dockerfile (not served here)
- benchmarks/ann-benchmarks/warm/config.yml (not served here)
- benchmarks/ann-benchmarks/warm/module.py (not served here)
- benchmarks/competitive/GAPS.json (not served here)
- benchmarks/competitive/README-block.md (not served here)
- benchmarks/competitive/ann_updatable_measure.py (not served here)
- benchmarks/competitive/lucene/LuceneBm25.java (not served here)
- benchmarks/competitive/modal_run.py (not served here)
- benchmarks/competitive/predictions/heavy-20261021T093821.json (not served here)
- benchmarks/competitive/repin_compare.py (not served here)
- benchmarks/competitive/requirements/annoy.footprint.in (not served here)
- benchmarks/competitive/requirements/annoy.footprint.txt (not served here)
- benchmarks/competitive/requirements/annoy.in (not served here)
- benchmarks/competitive/requirements/annoy.txt (not served here)
- benchmarks/competitive/requirements/bm25s.footprint.in (not served here)
- benchmarks/competitive/requirements/bm25s.footprint.txt (not served here)
- benchmarks/competitive/requirements/bm25s.in (not served here)
- benchmarks/competitive/requirements/bm25s.txt (not served here)
- benchmarks/competitive/requirements/chroma.footprint.in (not served here)
- benchmarks/competitive/requirements/chroma.footprint.txt (not served here)
- benchmarks/competitive/requirements/chroma.in (not served here)
- benchmarks/competitive/requirements/chroma.txt (not served here)
- benchmarks/competitive/requirements/diskann.footprint.in (not served here)
- benchmarks/competitive/requirements/diskann.footprint.txt (not served here)
- benchmarks/competitive/requirements/diskann.in (not served here)
- benchmarks/competitive/requirements/diskann.txt (not served here)
- benchmarks/competitive/requirements/duckdb.footprint.in (not served here)
- benchmarks/competitive/requirements/duckdb.footprint.txt (not served here)
- benchmarks/competitive/requirements/duckdb.in (not served here)
- benchmarks/competitive/requirements/duckdb.txt (not served here)
- benchmarks/competitive/requirements/faiss.footprint.in (not served here)
- benchmarks/competitive/requirements/faiss.footprint.txt (not served here)
- benchmarks/competitive/requirements/faiss.in (not served here)
- benchmarks/competitive/requirements/faiss.txt (not served here)
- benchmarks/competitive/requirements/hnswlib.footprint.in (not served here)
- benchmarks/competitive/requirements/hnswlib.footprint.txt (not served here)
- benchmarks/competitive/requirements/hnswlib.in (not served here)
- benchmarks/competitive/requirements/hnswlib.txt (not served here)
- benchmarks/competitive/requirements/lancedb.footprint.in (not served here)
- benchmarks/competitive/requirements/lancedb.footprint.txt (not served here)
- benchmarks/competitive/requirements/lancedb.in (not served here)
- benchmarks/competitive/requirements/lancedb.txt (not served here)
- benchmarks/competitive/requirements/milvus.footprint.in (not served here)
- benchmarks/competitive/requirements/milvus.footprint.txt (not served here)
- benchmarks/competitive/requirements/milvus.in (not served here)
- benchmarks/competitive/requirements/milvus.txt (not served here)
- benchmarks/competitive/requirements/ngt.footprint.in (not served here)
- benchmarks/competitive/requirements/ngt.footprint.txt (not served here)
- benchmarks/competitive/requirements/ngt.in (not served here)
- benchmarks/competitive/requirements/ngt.txt (not served here)
- benchmarks/competitive/requirements/pynndescent.footprint.in (not served here)
- benchmarks/competitive/requirements/pynndescent.footprint.txt (not served here)
- benchmarks/competitive/requirements/pynndescent.in (not served here)
- benchmarks/competitive/requirements/pynndescent.txt (not served here)
- benchmarks/competitive/requirements/qdrant.footprint.in (not served here)
- benchmarks/competitive/requirements/qdrant.footprint.txt (not served here)
- benchmarks/competitive/requirements/qdrant.in (not served here)
- benchmarks/competitive/requirements/qdrant.txt (not served here)
- benchmarks/competitive/requirements/scann.footprint.in (not served here)
- benchmarks/competitive/requirements/scann.footprint.txt (not served here)
- benchmarks/competitive/requirements/scann.in (not served here)
- benchmarks/competitive/requirements/scann.txt (not served here)
- benchmarks/competitive/requirements/server-clients.footprint.in (not served here)
- benchmarks/competitive/requirements/server-clients.footprint.txt (not served here)
- benchmarks/competitive/requirements/server-clients.in (not served here)
- benchmarks/competitive/requirements/server-clients.txt (not served here)
- benchmarks/competitive/requirements/sqlite-vec.footprint.in (not served here)
- benchmarks/competitive/requirements/sqlite-vec.footprint.txt (not served here)
- benchmarks/competitive/requirements/sqlite-vec.in (not served here)
- benchmarks/competitive/requirements/sqlite-vec.txt (not served here)
- benchmarks/competitive/requirements/stdlib-prev.footprint.in (not served here)
- benchmarks/competitive/requirements/stdlib-prev.footprint.txt (not served here)
- benchmarks/competitive/requirements/stdlib-prev.in (not served here)
- benchmarks/competitive/requirements/stdlib-prev.txt (not served here)
- benchmarks/competitive/requirements/stdlib.footprint.in (not served here)
- benchmarks/competitive/requirements/stdlib.footprint.txt (not served here)
- benchmarks/competitive/requirements/stdlib.in (not served here)
- benchmarks/competitive/requirements/stdlib.txt (not served here)
- benchmarks/competitive/requirements/tantivy-prev.footprint.in (not served here)
- benchmarks/competitive/requirements/tantivy-prev.footprint.txt (not served here)
- benchmarks/competitive/requirements/tantivy-prev.in (not served here)
- benchmarks/competitive/requirements/tantivy-prev.txt (not served here)
- benchmarks/competitive/requirements/tantivy.footprint.in (not served here)
- benchmarks/competitive/requirements/tantivy.footprint.txt (not served here)
- benchmarks/competitive/requirements/tantivy.in (not served here)
- benchmarks/competitive/requirements/tantivy.txt (not served here)
- benchmarks/competitive/requirements/usearch.footprint.in (not served here)
- benchmarks/competitive/requirements/usearch.footprint.txt (not served here)
- benchmarks/competitive/requirements/usearch.in (not served here)
- benchmarks/competitive/requirements/usearch.txt (not served here)
- benchmarks/competitive/requirements/voyager.footprint.in (not served here)
- benchmarks/competitive/requirements/voyager.footprint.txt (not served here)
- benchmarks/competitive/requirements/voyager.in (not served here)
- benchmarks/competitive/requirements/voyager.txt (not served here)
- benchmarks/competitive/requirements/whoosh.footprint.in (not served here)
- benchmarks/competitive/requirements/whoosh.footprint.txt (not served here)
- benchmarks/competitive/requirements/whoosh.in (not served here)
- benchmarks/competitive/requirements/whoosh.txt (not served here)
- benchmarks/competitive/tests/test_ann_systems.py (not served here)
- benchmarks/competitive/tests/test_common.py (not served here)
- benchmarks/competitive/tests/test_server.py (not served here)
- benchmarks/competitive/tests/test_table.py (not served here)
- benchmarks/competitive/tests/test_text_systems.py (not served here)
- benchmarks/competitive/tests/test_warm_ann_build.py (not served here)
- benchmarks/competitive/warm-bench (not served here)
- benchmarks/competitive/warmbench/__init__.py (not served here)
- benchmarks/competitive/warmbench/_spawner.py (not served here)
- benchmarks/competitive/warmbench/analyze.py (not served here)
- benchmarks/competitive/warmbench/analyzer_latency.py (not served here)
- benchmarks/competitive/warmbench/ann_worker.py (not served here)
- benchmarks/competitive/warmbench/cli.py (not served here)
- benchmarks/competitive/warmbench/common.py (not served here)
- benchmarks/competitive/warmbench/data.py (not served here)
- benchmarks/competitive/warmbench/envs.py (not served here)
- benchmarks/competitive/warmbench/h5_convert.py (not served here)
- benchmarks/competitive/warmbench/jars.py (not served here)
- benchmarks/competitive/warmbench/land.py (not served here)
- benchmarks/competitive/warmbench/multilingual.py (not served here)
- benchmarks/competitive/warmbench/registry.py (not served here)
- benchmarks/competitive/warmbench/runner.py (not served here)
- benchmarks/competitive/warmbench/server_proc.py (not served here)
- benchmarks/competitive/warmbench/server_registry.py (not served here)
- benchmarks/competitive/warmbench/server_runner.py (not served here)
- benchmarks/competitive/warmbench/server_worker.py (not served here)
- benchmarks/competitive/warmbench/servers.py (not served here)
- benchmarks/competitive/warmbench/table.py (not served here)
- benchmarks/competitive/warmbench/text_worker.py (not served here)
- benchmarks/competitive/warmbench/textdata.py (not served here)
- benchmarks/cue_tags/cue_tagger.py (not served here)
- benchmarks/cue_tags/labels.csv (not served here)
- benchmarks/cue_tags/measure.py (not served here)
- benchmarks/cue_tags/sample.py (not served here)
- benchmarks/cue_tags/stale_arms.py (not served here)
- benchmarks/embedding/agreement.py (not served here)
- benchmarks/embedding/agreement_sample.json (not served here)
- benchmarks/embedding/assemble_selection.py (not served here)
- benchmarks/embedding/bench_torch_cpu.py (not served here)
- benchmarks/embedding/embed_store_bench.rs (not served here)
- benchmarks/embedding/pin_model.py (not served here)
- benchmarks/embedding/reference_granite-embedding-30m-english.json (not served here)
- benchmarks/embedding/run_engine_eval.py (not served here)
- benchmarks/embedding/run_store_eval.py (not served here)
- benchmarks/embedding/select_model.py (not served here)
- benchmarks/evidence/MANIFEST.json (not served here)
- benchmarks/evidence/README.md (not served here)
- benchmarks/evidence/retired-inputs.json (not served here)
- benchmarks/modal_benchmark.py (not served here)
- benchmarks/private_search/lattice_estimate.py (not served here)
- benchmarks/private_search/lwe_core_svp.py (not served here)
- benchmarks/public/accuracy_qualification/PLAN.json (not served here)
- benchmarks/public/accuracy_qualification/analyze.mjs (not served here)
- benchmarks/public/accuracy_qualification/prepare.py (not served here)
- benchmarks/public/accuracy_qualification/pyproject.toml (not served here)
- benchmarks/public/accuracy_qualification/run.mjs (not served here)
- benchmarks/public/accuracy_qualification/test-analyze.mjs (not served here)
- benchmarks/public/accuracy_qualification/test_prepare.py (not served here)
- benchmarks/public/accuracy_qualification/uv.lock (not served here)
- benchmarks/public/aml_cycle2/README.md (not served here)
- benchmarks/public/aml_cycle2/adapter.py (not served here)
- benchmarks/public/aml_cycle2/experimental/__init__.py (not served here)
- benchmarks/public/aml_cycle2/experimental/citation_matches.csv (not served here)
- benchmarks/public/aml_cycle2/experimental/evidence_selection.py (not served here)
- benchmarks/public/aml_cycle2/experimental/findings.json (not served here)
- benchmarks/public/aml_cycle2/experimental/semantic_cases.json (not served here)
- benchmarks/public/aml_cycle2/experimental/test_evidence_selection.py (not served here)
- benchmarks/public/aml_cycle2/gce_install.sh (not served here)
- benchmarks/public/aml_cycle2/local_host.py (not served here)
- benchmarks/public/aml_cycle2/modal_deploy.py (not served here)
- benchmarks/public/aml_cycle2/runtime-requirements.in (not served here)
- benchmarks/public/aml_cycle2/runtime-requirements.txt (not served here)
- benchmarks/public/aml_cycle2/test_adapter.py (not served here)
- benchmarks/public/aml_cycle2/test_local_host.py (not served here)
- benchmarks/public/aml_cycle2/test_query.py (not served here)
- benchmarks/public/analyze_glove_batch_pool.py (not served here)
- benchmarks/public/analyze_glove_batch_pool_feature.py (not served here)
- benchmarks/public/analyze_glove_batch_pool_isolation.py (not served here)
- benchmarks/public/analyze_glove_development.py (not served here)
- benchmarks/public/ann_extra.py (not served here)
- benchmarks/public/ann_failure_map.py (not served here)
- benchmarks/public/ann_modal_confirm.py (not served here)
- benchmarks/public/ann_modal_graph.py (not served here)
- benchmarks/public/ann_modal_repair.py (not served here)
- benchmarks/public/audit_calibration.py (not served here)
- benchmarks/public/audit_serving_job.py (not served here)
- benchmarks/public/batching_assemble.py (not served here)
- benchmarks/public/batching_compare.py (not served here)
- benchmarks/public/batching_inputs.py (not served here)
- benchmarks/public/batching_timing.py (not served here)
- benchmarks/public/bemula/README.md (not served here)
- benchmarks/public/bemula/run_warm.py (not served here)
- benchmarks/public/bemula/summarize.py (not served here)
- benchmarks/public/bemula/warm_adapter.py (not served here)
- benchmarks/public/build_native_bridge.py (not served here)
- benchmarks/public/check_algebraic_codebook_report.py (not served here)
- benchmarks/public/check_compaction_updates.py (not served here)
- benchmarks/public/check_competitive_ann.py (not served here)
- benchmarks/public/check_document_competition.py (not served here)
- benchmarks/public/check_document_operations.py (not served here)
- benchmarks/public/check_document_search_compare.py (not served here)
- benchmarks/public/check_document_sqlite_rank_control.py (not served here)
- benchmarks/public/check_encoder_version_diagnostic.py (not served here)
- benchmarks/public/check_evidence_budget.py (not served here)
- benchmarks/public/check_evidence_completeness.py (not served here)
- benchmarks/public/check_factored_evidence.py (not served here)
- benchmarks/public/check_geometry_report.py (not served here)
- benchmarks/public/check_learned_sparse.py (not served here)
- benchmarks/public/check_lexical_relevance.py (not served here)
- benchmarks/public/check_longmemeval_answers.py (not served here)
- benchmarks/public/check_longmemeval_complete_history.py (not served here)
- benchmarks/public/check_longmemeval_full_history.py (not served here)
- benchmarks/public/check_longmemeval_production.py (not served here)
- benchmarks/public/check_longmemeval_role_reader.py (not served here)
- benchmarks/public/check_longmemeval_user_first.py (not served here)
- benchmarks/public/check_longmemeval_verifier.py (not served here)
- benchmarks/public/check_manifold_report.py (not served here)
- benchmarks/public/check_manifold_schema_report.py (not served here)
- benchmarks/public/check_native_harness.py (not served here)
- benchmarks/public/check_operand_report.py (not served here)
- benchmarks/public/check_operand_tampering.py (not served here)
- benchmarks/public/check_phase_path_report.py (not served here)
- benchmarks/public/check_publication_serving.py (not served here)
- benchmarks/public/check_reader_suspensions.py (not served here)
- benchmarks/public/check_sparse_composition.py (not served here)
- benchmarks/public/check_superiority_checkpoint.py (not served here)
- benchmarks/public/check_weak_reader_evidence.py (not served here)
- benchmarks/public/compact_owned_publication.py (not served here)
- benchmarks/public/compaction_updates.py (not served here)
- benchmarks/public/compare_fact_caches.py (not served here)
- benchmarks/public/competitive_ann.py (not served here)
- benchmarks/public/direct_commit_publication.py (not served here)
- benchmarks/public/document_competition.py (not served here)
- benchmarks/public/document_search_compare.py (not served here)
- benchmarks/public/document_sqlite_rank_control.py (not served here)
- benchmarks/public/durability_matrix.py (not served here)
- benchmarks/public/encoder_version_diagnostic.py (not served here)
- benchmarks/public/evaluate.py (not served here)
- benchmarks/public/evidence_budget_diagnostic.py (not served here)
- benchmarks/public/evidence_completeness.py (not served here)
- benchmarks/public/factored_evidence_diagnostic.py (not served here)
- benchmarks/public/geometry_experiments.py (not served here)
- benchmarks/public/instrument_publication.py (not served here)
- benchmarks/public/learned_sparse_dense_tune.py (not served here)
- benchmarks/public/learned_sparse_encode.py (not served here)
- benchmarks/public/learned_sparse_results.py (not served here)
- benchmarks/public/lexical_relevance_tune.py (not served here)
- benchmarks/public/lme_context/README.md (not served here)
- benchmarks/public/lme_context/breakdown.py (not served here)
- benchmarks/public/lme_context/extract_modal.py (not served here)
- benchmarks/public/lme_context/prep_extract.py (not served here)
- benchmarks/public/lme_context/render.py (not served here)
- benchmarks/public/lme_context/run_arm.py (not served here)
- benchmarks/public/locomo/assemble_results.py (not served here)
- benchmarks/public/locomo/locomo_bench.py (not served here)
- benchmarks/public/locomo/mem0_prompts.py (not served here)
- benchmarks/public/locomo/test_locomo_bench.py (not served here)
- benchmarks/public/locomo_comparison.py (not served here)
- benchmarks/public/longmemeval_500_frozen_config.json (not served here)
- benchmarks/public/longmemeval_answer_evidence.py (not served here)
- benchmarks/public/longmemeval_compact_context.py (not served here)
- benchmarks/public/longmemeval_complete_history.py (not served here)
- benchmarks/public/longmemeval_config.json (not served here)
- benchmarks/public/longmemeval_dev_conversations.py (not served here)
- benchmarks/public/longmemeval_dev_sources.py (not served here)
- benchmarks/public/longmemeval_full_history.py (not served here)
- benchmarks/public/longmemeval_history_budget.py (not served here)
- benchmarks/public/longmemeval_incremental_judges.py (not served here)
- benchmarks/public/longmemeval_local_modal.py (not served here)
- benchmarks/public/longmemeval_local_reader.py (not served here)
- benchmarks/public/longmemeval_memory_recall.py (not served here)
- benchmarks/public/longmemeval_modal.py (not served here)
- benchmarks/public/longmemeval_obligations.py (not served here)
- benchmarks/public/longmemeval_official_qa.py (not served here)
- benchmarks/public/longmemeval_official_qa_500.py (not served here)
- benchmarks/public/longmemeval_pipeline.py (not served here)
- benchmarks/public/longmemeval_production.py (not served here)
- benchmarks/public/longmemeval_qa_l01.py (not served here)
- benchmarks/public/longmemeval_quality.py (not served here)
- benchmarks/public/longmemeval_quota.py (not served here)
- benchmarks/public/longmemeval_reader_api.py (not served here)
- benchmarks/public/longmemeval_reader_api_v2.py (not served here)
- benchmarks/public/longmemeval_reader_budget.py (not served here)
- benchmarks/public/longmemeval_retrieval.py (not served here)
- benchmarks/public/longmemeval_retrieval_limits.py (not served here)
- benchmarks/public/longmemeval_role_reader.py (not served here)
- benchmarks/public/longmemeval_role_retrieval.py (not served here)
- benchmarks/public/longmemeval_run_preopen.sh (not served here)
- benchmarks/public/longmemeval_run_preopen_r3.sh (not served here)
- benchmarks/public/longmemeval_session_closure.py (not served here)
- benchmarks/public/longmemeval_sota.py (not served here)
- benchmarks/public/longmemeval_sota_closure.py (not served here)
- benchmarks/public/longmemeval_sota_context.py (not served here)
- benchmarks/public/longmemeval_sota_expand.py (not served here)
- benchmarks/public/longmemeval_sota_hybrid.py (not served here)
- benchmarks/public/longmemeval_sota_receipts.py (not served here)
- benchmarks/public/longmemeval_sota_replay.py (not served here)
- benchmarks/public/longmemeval_sota_report.py (not served here)
- benchmarks/public/longmemeval_split.json (not served here)
- benchmarks/public/longmemeval_split.py (not served here)
- benchmarks/public/longmemeval_sweep.py (not served here)
- benchmarks/public/longmemeval_user_first.py (not served here)
- benchmarks/public/longmemeval_verifier.py (not served here)
- benchmarks/public/longmemeval_verifier_sources.py (not served here)
- benchmarks/public/longmemeval_weak_reader.py (not served here)
- benchmarks/public/lucene_hnsw/LuceneHnsw.java (not served here)
- benchmarks/public/matched_server/PLAN.json (not served here)
- benchmarks/public/matched_server/adapters.py (not served here)
- benchmarks/public/matched_server/calibrate.py (not served here)
- benchmarks/public/matched_server/check_adapters.py (not served here)
- benchmarks/public/matched_server/check_measure.py (not served here)
- benchmarks/public/matched_server/check_qualification.py (not served here)
- benchmarks/public/matched_server/claim_gate.py (not served here)
- benchmarks/public/matched_server/diagnose_search.py (not served here)
- benchmarks/public/matched_server/distribution.py (not served here)
- benchmarks/public/matched_server/measure.py (not served here)
- benchmarks/public/matched_server/oracle.py (not served here)
- benchmarks/public/matched_server/prepare_inputs.py (not served here)
- benchmarks/public/matched_server/pyproject.toml (not served here)
- benchmarks/public/matched_server/qualify.py (not served here)
- benchmarks/public/matched_server/seal_qualification.py (not served here)
- benchmarks/public/matched_server/test_claim_gate.py (not served here)
- benchmarks/public/matched_server/test_distribution.py (not served here)
- benchmarks/public/matched_server/test_measure.py (not served here)
- benchmarks/public/matched_server/test_prepare_inputs.py (not served here)
- benchmarks/public/matched_server/test_protocol.py (not served here)
- benchmarks/public/matched_server/test_qualify.py (not served here)
- benchmarks/public/matched_server/test_seal_qualification.py (not served here)
- benchmarks/public/matched_server/uv.lock (not served here)
- benchmarks/public/matched_server/verify_evidence.py (not served here)
- benchmarks/public/mem0_protocol/README.md (not served here)
- benchmarks/public/mem0_protocol/budget_ledger.py (not served here)
- benchmarks/public/mem0_protocol/clear_failed.py (not served here)
- benchmarks/public/mem0_protocol/grade_validation.py (not served here)
- benchmarks/public/mem0_protocol/live_smoke.py (not served here)
- benchmarks/public/mem0_protocol/manifest_template.json (not served here)
- benchmarks/public/mem0_protocol/openai_proxy.py (not served here)
- benchmarks/public/mem0_protocol/price.py (not served here)
- benchmarks/public/mem0_protocol/protocol_scope.py (not served here)
- benchmarks/public/mem0_protocol/quota_resume.py (not served here)
- benchmarks/public/mem0_protocol/retrieval_config.json (not served here)
- benchmarks/public/mem0_protocol/retrieval_config_count_matched.json (not served here)
- benchmarks/public/mem0_protocol/retrieval_config_hybrid_dev.json (not served here)
- benchmarks/public/mem0_protocol/retrieval_config_ordered_dev.json (not served here)
- benchmarks/public/mem0_protocol/reuse_locomo_dev.py (not served here)
- benchmarks/public/mem0_protocol/run_mem0_protocol.py (not served here)
- benchmarks/public/mem0_protocol/shim.py (not served here)
- benchmarks/public/mem0_protocol/test_budget_ledger.py (not served here)
- benchmarks/public/mem0_protocol/test_clear_failed.py (not served here)
- benchmarks/public/mem0_protocol/test_openai_proxy.py (not served here)
- benchmarks/public/mem0_protocol/test_quota_resume.py (not served here)
- benchmarks/public/mem0_protocol/test_reuse_locomo_dev.py (not served here)
- benchmarks/public/mem0_protocol/test_run_mem0_protocol.py (not served here)
- benchmarks/public/mem0_protocol/test_shim.py (not served here)
- benchmarks/public/mobile_qualification/ANDROID-PLAN.json (not served here)
- benchmarks/public/mobile_qualification/IOS-PLAN.json (not served here)
- benchmarks/public/mobile_qualification/PLAN.json (not served here)
- benchmarks/public/mobile_qualification/RecoveryConsumerTest.kt (not served here)
- benchmarks/public/native_bridge/Cargo.toml (not served here)
- benchmarks/public/native_bridge/build.rs (not served here)
- benchmarks/public/native_bridge/src/lib.rs (not served here)
- benchmarks/public/native_bridge/src/vertex_support.rs (not served here)
- benchmarks/public/native_harness.py (not served here)
- benchmarks/public/native_recovery/PLAN.json (not served here)
- benchmarks/public/native_recovery/Program.cs (not served here)
- benchmarks/public/native_recovery/RecoveryConsumer.java (not served here)
- benchmarks/public/native_recovery/recovery.c (not served here)
- benchmarks/public/native_recovery/recovery.go (not served here)
- benchmarks/public/native_recovery/run.py (not served here)
- benchmarks/public/native_recovery/test_protocol.py (not served here)
- benchmarks/public/node_recovery.cjs (not served here)
- benchmarks/public/operand_experiment.py (not served here)
- benchmarks/public/platform_microbench_aggregate.py (not served here)
- benchmarks/public/platform_modal_avx512.py (not served here)
- benchmarks/public/platform_modal_e2e.py (not served here)
- benchmarks/public/platform_paired.py (not served here)
- benchmarks/public/prepare.py (not served here)
- benchmarks/public/prepare_ann_scale.py (not served here)
- benchmarks/public/proxies/ann_proxies.py (not served here)
- benchmarks/public/proxies/beir_proxies.py (not served here)
- benchmarks/public/proxies/lme_proxies.py (not served here)
- benchmarks/public/proxies/qgraph_sim.py (not served here)
- benchmarks/public/proxies/stop_screen_report.py (not served here)
- benchmarks/public/publication_serving.py (not served here)
- benchmarks/public/publication_serving.rs (not served here)
- benchmarks/public/publication_trace.rs (not served here)
- benchmarks/public/python_sdk_recovery.py (not served here)
- benchmarks/public/qag_serving_execute.py (not served here)
- benchmarks/public/qag_serving_module.py (not served here)
- benchmarks/public/qgraph_paired.py (not served here)
- benchmarks/public/qgraph_table.py (not served here)
- benchmarks/public/reader_protocol_v1.json (not served here)
- benchmarks/public/remote_document_pilot.py (not served here)
- benchmarks/public/replay_capacity_publication.py (not served here)
- benchmarks/public/report_sparse_composition.py (not served here)
- benchmarks/public/restore_ann_checkpoint.py (not served here)
- benchmarks/public/resume_longmemeval_readers.py (not served here)
- benchmarks/public/run_glove_batch_pool_ablation.py (not served here)
- benchmarks/public/run_glove_batch_pool_feature.py (not served here)
- benchmarks/public/run_glove_batch_pool_isolation.py (not served here)
- benchmarks/public/run_glove_candidate_checks.py (not served here)
- benchmarks/public/run_glove_development.py (not served here)
- bindings/js/warm-sdk/LICENSE
- docs/ARCHITECTURE.md
- docs/COMPARISON.md
- docs/DETERMINISTIC-RESONATOR.md
- docs/GOVERNANCE.md
- docs/INTEGRATION.md
- docs/PREREGISTRATION-binding-readout.md
- docs/PRIVACY.md
- docs/PRIVATE-SEARCH.md
- docs/PUBLIC-BENCHMARKS.md
- docs/RELEASING.md
- docs/RESEARCH.md
- docs/SECURITY.md
- docs/SERVER.md
- docs/STABILITY.md
- docs/SUPERPOSITION-FLOOR.md
- docs/assets/fonts/LICENSE-OFL.txt
- docs/evaluation.md
- docs/index.md
- docs/production-readiness.md
- integrations/mcp/npm/LICENSE
- integrations/mcp/pypi/LICENSE
- tests/fixtures/vectra-0.15/LICENSE
Advantages
- Supports local-first operation with no external service calls or telemetry. evidence: README.md
- Provides a memory API that allows for explicit fact removal via a forget operation. evidence: README.md
Limitations
- The system supports only one writer per store. evidence: README.md
- The vector index build time and serving RSS are currently behind hnswlib on SIFT-1M benchmarks. evidence: README.md
Security
Risk level: Medium risk
This brain runs
Install scripts (1)
benchmarks/results/sbg_positions_development/setup.sh:sh benchmarks/results/sbg_positions_development/setup.sh
MCP servers (3)
integrations/editors/claude-code/plugin/.mcp.json(writerslogic-warm):npx -y writerslogic-warm-mcpintegrations/editors/claude-desktop/claude_desktop_config.json(writerslogic-warm):npx -y writerslogic-warm-mcpintegrations/editors/gemini-cli/gemini-extension.json(writerslogic-warm):npx -y writerslogic-warm-mcp
Hooks (1)
integrations/editors/claude-code/plugin/hooks/hooks.json(SessionStart):node
- Medium risk: runs-mcp-server-unpinned-runner (integrations/editors/claude-code/plugin/.mcp.json)
- Medium risk: runs-mcp-server-unpinned-runner (integrations/editors/claude-desktop/claude_desktop_config.json)
- Medium risk: runs-mcp-server-unpinned-runner (integrations/editors/gemini-cli/gemini-extension.json)
- Low risk: scripts-present (benchmarks/results/sbg_positions_development/setup.sh)
Attribution and licence
- Created by
- Writerslogic
- Declared authors
- WritersLogic, Inc. (declared in package.json)
- WritersLogic, Inc. (declared in pyproject.toml)
- WritersLogic Contributors (declared in CITATION.cff)
- Publisher
- Writerslogic (Official Writerslogic project)
- Source
- writerslogic/warm · commit e176d55
- Licence
- Apache License 2.0 (files served here)
- ATTRIBUTION.md (written by AutoGineer)
- Synced
AutoGineer is not affiliated with or endorsed by Writerslogic.
Install from source
481 of its files are not served here, so there is no ZIP. writerslogic/warm at e176d55