{ "markdown": "\n

\n \"VelesDB\"\n

\n

VelesDB

\n

\n One ~14 MB binary fuses vector + graph + columnar under a single query language — with an agent memory that shows its evidence and a deterministic context compiler that cuts your real, billed token spend.
\n Local-first: nothing leaves the machine, no LLM and no API key in the memory path. Every number below links to a committed harness you can rerun.\n

\n

\n \"CI\" \"Crates.io\" \"PyPI\" \"npm\" \"Codacy \"License\"
\n Quick start • Proof • Limitations • Architecture • Roadmap • velesdb.com\n

\n\n---\n\n\n\n## Start here — three commands that work\n\n```bash\npip install velesdb\ncurl -O https://raw.githubusercontent.com/cyberlife-coder/VelesDB/main/examples/python/hello_velesdb.py\npython hello_velesdb.py\n```\n\nTo search your own text instead of hand-written vectors, install the opt-in local adapter (`pip install \"velesdb[embed-sentence-transformers]\"`) and run [`hello_velesdb_text.py`](examples/python/hello_velesdb_text.py); its first run downloads `all-MiniLM-L6-v2`.\n\nExpected output, byte-for-byte ([read the script](examples/python/hello_velesdb.py) — no server, no embedding model):\n\n```\nQuery: \"tech\"\n score=1.000 Rust 1.89 release notes\n score=0.600 AI-generated jazz: the new wave\n score=0.000 Best ramen in Tokyo\n\nQuery: \"tech + music\"\n score=0.990 AI-generated jazz: the new wave\n score=0.707 Rust 1.89 release notes\n score=0.707 Miles Davis discography\n```\n\nAn embedding model determines the vector dimension, while your similarity semantics determine the metric. Both are fixed when a collection is created; to change either, create a new collection and re-index your documents.\n\n**Give your agent a persistent memory — three more commands:**\n\n```bash\ncargo install velesdb-memory # the local MCP memory server\nclaude mcp add velesdb-memory -- ~/.cargo/bin/velesdb-memory # any MCP client works\ncurl -L https://github.com/cyberlife-coder/VelesDB/releases/latest/download/velesdb-skills.tar.gz | tar -xz -C ~/.claude/skills/\n```\n\nNo Rust toolchain? `npm i @wiscale/velesdb-memory-node`, or grab a prebuilt `.mcpb` bundle from the [official MCP Registry](https://registry.modelcontextprotocol.io/?q=velesdb-memory) (`io.github.cyberlife-coder/velesdb-memory`).\n\n
\nOther paths — always-on hooks, shared daemon, Rust, Docker, WASM, REST\n\n**Memory used *continuously*, not just available:** [`integrations/agent-hooks/`](integrations/agent-hooks/README.md) wires five Claude Code hooks — `SessionStart`/`Stop`/`PreCompact` resume and save the working context, `PreToolUse` requires successful recall before an opted-in repository edit, and `PostToolUse` both records that recall and compiles an oversized tool result *before* it enters the transcript. One global install covers every project without enabling the edit guard outside explicitly configured repositories.\n\n**One memory shared by several clients** (Claude Code, Codex CLI, Claude Desktop, Windsurf, Devin CLI): [`scripts/install-memory-daemon.sh`](crates/velesdb-memory/README.md#http-transport-multi-client) runs `velesdb-memory` as a single local daemon — HTTPS by default, with a natively generated local CA.\n\n**Cargo (Rust + REST server):** `cargo install velesdb-server velesdb-cli` — **Docker** (multi-arch linux/amd64 + linux/arm64): `docker run -d -p 8080:8080 -v velesdb_data:/data --name velesdb ghcr.io/cyberlife-coder/velesdb:latest`, then `curl http://localhost:8080/health`.\n\n**Browser / edge:** the WASM build is ~710 KB gzipped and runs entirely client-side ([TypeScript SDK](sdks/typescript)). **REST:** 54 REST endpoints ([OpenAPI spec](docs/openapi.yaml)). Full matrix: [installation guide](docs/guides/INSTALLATION.md).\n\n
\n\n---\n\n## Why VelesDB\n\n- **One database instead of three.** Vectors for *\"what feels similar\"*, a graph for *\"what is connected\"*, typed columns for *\"what I know for sure\"* — normally three deployments, three query languages, and glue code. Here it is one binary and [one language](docs/VELESQL_SPEC.md).\n- **A memory that can be audited, not just queried.** Every recall can show the evidence behind it; every compression decision carries a rule id, a reason, and a risk level. Deterministic by construction — no model in the write path, so no drift and nothing to re-litigate.\n- **Local-first is a sovereignty decision, not a latency one.** No cloud, no API key, no data processor: air-gapped if you want it, in your jurisdiction by default. [Why that matters](https://dev.to/wiscale-fr/i-built-a-database-in-france-because-the-cloud-act-makes-eu-data-sovereignty-impossible-5325) · [positioning in depth](docs/WHY_VELESDB.md).\n\n## How it works, in plain terms\n\nFour things happen, and none of them calls an AI provider.\n\n**1 · It stores facts, not conversations.** You give it one statement — *\"the\nAPI port is 6333 because 3000 collided with the web UI\"* — and it lands in a\nlocal file store. No model call, nothing sent anywhere.\n\n**2 · It finds them by meaning.** Asking *\"which port did we settle on\"* reaches\nthat fact even though none of the words match. A local embedding model turns\ntext into coordinates; close meaning means close coordinates.\n\n**3 · It connects them, and that is the part a search engine cannot do.** Each\nfact is linked to the topics it mentions. `why()` starts from the best match and\nthen **walks those links**, so it returns the answer *plus the facts that\nexplain it* — including ones sharing no vocabulary with your question.\n\n> The links have to exist. Store facts one by one and the graph stays flat, so\n> `why()` behaves like a search. Hand a paragraph to `remember_extracted` and it\n> splits it into facts and wires the links for you.\n\n**4 · It compresses what is too big, before you pay for it.** Give the compiler\nyour accumulated context and a token budget; it returns a smaller version with\n**one recorded decision per fragment** — kept, abstracted, or dropped — and a\nhandle to fetch any original back verbatim. Same input, same bytes out, every\ntime. That is what the [82.5 % below](#proof--three-numbers-each-tied-to-its-harness) measures.\n\n---\n\n## What no one else combines\n\n### 1 · Three engines, one query\n\n| Engine | What it does |\n|---|---|\n| **Vector** | Semantic similarity (HNSW + AVX2/NEON SIMD) |\n| **Graph** | Typed relationships, BFS/DFS, native `MATCH` clause ([patterns](docs/guides/GRAPH_PATTERNS.md)) |\n| **ColumnStore** | Typed columnar metadata filtering, secondary indexes |\n\nOne statement crosses all three — similarity, relations and typed filters, no glue code:\n\n```sql\nMATCH (doc:Document)-[:AUTHORED_BY]->(author:Person)\nWHERE similarity(doc.embedding, $question) > 0.8\n AND author.department = 'Engineering'\nRETURN author.name, doc.title\nORDER BY similarity() DESC LIMIT 5\n```\n\n### 2 · A memory that shows its evidence — `why()`\n\nMost \"agent memory\" is vector recall: it finds text that *looks like* your query. VelesDB connects memories with typed links, so it can answer *why* something happened by walking the graph to context that shares **no words** with your question — across process restarts, offline, no API key:\n\n```python\nfrom velesdb import MemoryService # pip install velesdb\n\nmem = MemoryService(\"./agent_memory\") # a real on-disk store; survives restarts\nreason = mem.remember(\"Robert is recovering from knee surgery\")\nmem.remember(\"Booked the aisle seat on Robert's flight\", links=[(reason, \"because\")])\n\n# A *new* process, weeks later, reopens the same store and asks why:\nmem.why(\"why the aisle seat on Robert's flight?\") # walks booking → reason — recall() can't\n```\n\n> `MemoryService` defaults to the offline `hash` embedder: deterministic, but\n> **lexical rather than semantic**, so unrelated wording can score `0.000`.\n> Opening it now says so once on stderr. For meaning-based recall, pass\n> `embedder=\"ollama\"`; follow [Real semantic recall in 5 minutes](crates/velesdb-memory/README.md#real-semantic-recall-in-5-minutes).\n\n![recall() finds the booking but misses the reason; why() reaches it through typed links, across a session restart](examples/agent_memory/why_across_sessions.gif)\n\nMemories are permanent by default; `forget(id)` deletes one, `ttl_seconds` gives a fact a durable expiry. Every `remember` auto-stamps its storage day, so recency-weighted recall works with zero setup. Same wedge in **Python**, **Node**, the [**MCP server**](crates/velesdb-memory), and in-memory in the [**TypeScript SDK**](sdks/typescript).\n\nProof it is not a weak-embedder trick — four runnable demos in which `recall` stays blind to the reason **even under a real semantic embedder** (`ollama` / `all-minilm`), because the reason is connected by a decision rather than by surface similarity: [`why_across_sessions.py`](examples/agent_memory/why_across_sessions.py) (survives a process restart) · [`why_magic_constant.py`](examples/agent_memory/why_magic_constant.py) (a business reason sharing no words with the code) · [`memory_builds_its_own_graph.py`](examples/agent_memory/memory_builds_its_own_graph.py) (raw prose in, auto-wired graph out) · [`why_magic_constant.mjs`](crates/velesdb-node/examples/why_magic_constant.mjs) (Node). Benchmark position, including LoCoMo and why cross-lab scores are not fairly comparable: [BENCHMARK.md](crates/velesdb-memory/BENCHMARK.md) · [Agent Memory guide](docs/guides/AGENT_MEMORY.md).\n\n### 3 · A deterministic context compiler\n\nAgents burn most of their budget re-reading redundant context. `compile_context` / `compile_transcript` (MCP, or `ContextCompiler` in Rust) shrink it with **no LLM and no network**:\n\n- **Deterministic** — the same input always compiles to the same bytes, asserted twice per run in every committed benchmark. That also yields a [byte-stable cache prefix](crates/velesdb-memory/examples/context_savings/real_measures/cache_prefix.mjs) provider prompt-caching can actually hit.\n- **Auditable** — `explain_compilation` gives every kept or dropped fragment a stable rule id, a reason and a risk level.\n- **Reversible** — over-budget content becomes a recoverable `ctx://source/` handle; `retrieve_context_source` brings the original bytes back on demand.\n- **Bounded** — it compresses only what your agent explicitly hands it, never the harness's system prompt, and nothing enters recallable memory without an explicit `remember`.\n\nCode, URLs, numbers and negative constraints survive verbatim. The [`velesdb-context-optimizer` skill](skills/velesdb-context-optimizer/SKILL.md) teaches the workflow — including when *not* to compress.\n\n---\n\n## Proof — three numbers, each tied to its harness\n\nNo figure here is an estimate from a slide; each links to the log or script in this repo that produced it.\n\n| Claim | Measured | Harness |\n|---|---|---|\n| Real **billed dollars** saved, same agent session sent raw vs compiled (real Claude billing, deterministic fact-checklist grader — no LLM judge) | **21.9 %** at real Retina screenshot weight, quality at parity (23.0/23 facts both arms) | [real-session-benchmark](examples/real-session-benchmark#billed-campaign-results-2026-07-19-cli-runner-claude-sonnet-5) · [raw logs](examples/real-session-benchmark/results/2026-07-19-vibe-cli/) |\n| Real (cl100k) **input-token savings** on a committed 12-turn agent-session corpus | **82.5 %**, compiled in ~0.5 ms mean stateless (~27 ms with source persistence on) | [context_savings](crates/velesdb-memory/examples/context_savings) |\n| **Vector search** latency on the full production path (VelesQL → HNSW → WAL ON → payload hydration) | **450 us** p50 (10K/384D, recall ≥ 96 %) | [docs/BENCHMARKS.md](docs/BENCHMARKS.md) |\n\n> Same campaign, less flattering: **10.9 %** on cropped screenshots, **14.7 %** on a 36-turn day-scale arc, **15.1 %** input tokens on the direct Messages API, and **2.5 %** for the no-screenshots variant — that spread *is* the measured value of the media mechanisms, so we publish it as prominently as the headline. [Honest reading, limitations and full protocol](examples/real-session-benchmark#honest-limitations). Every number on this page is CI-guarded by a [promise contract](docs/reference/promise-contract.json) that pins the README to its committed sources.\n\n
\nThe full measurement tables — billed A/B runs, retrieval quality, engine micro-benchmarks\n\n**Billed A/B sessions** (2026-07-19, claude-sonnet-5; raw logs committed verbatim):\n\n| Session | Runner | $ saved | Quality (raw vs compiled) |\n|---|---|---|---|\n| 19-turn feature session, cropped screenshots | Claude CLI | **10.9 %** | 22.8/23 vs 23.0/23 facts |\n| Same session, real Retina-weight screenshots | Claude CLI | **21.9 %** | 23.0/23 vs 23.0/23 |\n| 36-turn day-scale session | Claude CLI | **14.7 %** | 49.6/50 vs 49.2/50 * |\n| 19-turn session, direct Messages API | API | **15.1 %** input tokens | 23.0/23 vs 23.0/23 |\n\n> \\* Two turns' grading key was later found defective (both arms scored full marks there; the parity conclusion stands) — [disclosure](examples/real-session-benchmark#billed-results-2026-07-19-all-real-executions). Over a [36-turn session](examples/real-session-benchmark#long-session-36-turns--context-window-headroom) compiled context grows **1.7× slower**, so one session lasts far longer before hitting the window.\n\n**Memory retrieval quality**, public test sets, no AI grader in the loop: **+7.2 pts** multi-hop (HotpotQA), **+9.7 pts** time-scoped recall (TimeQA), **+29 pts** on a controlled task needing both engines at once — [BENCHMARK.md](crates/velesdb-memory/BENCHMARK.md).\n\n**End-to-end search** (canonical): search p50 **450 us** (10K, 384D, WAL ON) · SIMD dot product **21.7 ns** (768D, AVX2) · Recall@10 balanced 98.8 % · quantization PQ (8–32x), RaBitQ (32x), SQ8 (4x), Binary (32x) — [scope & caveats](docs/guides/QUANTIZATION.md).\n\n**Index-only micro-benchmarks** (no WAL, no payload, hot cache — *not* comparable to the end-to-end figure above), each reproducible with `cargo bench -p velesdb-core --bench `: HNSW Search index-only (10K/768D, k=10) **55 us** (`hnsw_benchmark -- hnsw_search_latency`) · SIMD Dot Product (768D, AVX2) **21.7 ns** (`simd_benchmark`) · Recall@10 accurate mode **100%** (`recall_benchmark`) · BM25 Sparse Search index-only (10K docs, top-10) 57.6 us (`sparse_benchmark -- top10_10k_corpus`).\n\n| Search mode | ef_search | Recall@10 | Use case |\n|---|---|---|---|\n| Fast | 64 | 92.2% | Real-time suggestions, typeahead |\n| Balanced (default) | 128 | 98.8% | Production search, RAG pipelines |\n| Accurate | 512 | 100% | Evaluation, ground truth comparison |\n\n**Distance metrics** — 5 with SIMD acceleration (AVX-512, AVX2, NEON), at 768D/AVX2 on hot cache: Cosine 33 ns · Euclidean 20 ns · Dot Product 22 ns · Hamming 36 ns · Jaccard 35 ns.\n\n**ColumnStore** — typed columnar filtering, **130x faster** than JSON scanning at 100K rows on the i9-14900KF reference (`JSON scan 3.84 ms → ColumnStore 29.5 us`). The ratio is hardware-dependent: on Apple Silicon (M5 Pro, 2026-07-20) the JSON scan itself runs ~2.8× faster, so the same bench measures ~50–105x while the ColumnStore's absolute time holds (~27 µs).\n\n> **Provenance:** Intel Core **i9-14900KF** (x86_64, AVX2). Per-machine figures vary; Apple-Silicon cross-checks, the SIFT1M standardized ANN run and the full methodology live in [docs/BENCHMARKS.md](docs/BENCHMARKS.md). Reproduce the end-to-end figure with `python benchmarks/velesdb_benchmark.py --recall`.\n\n
\n\n---\n\n## Pick your entry point\n\n| I want to… | Use | Notes |\n|---|---|---|\n| Try it in one file | [`velesdb`](https://pypi.org/project/velesdb/) (Python 3.9+) | Fastest onboarding path |\n| Embed the engine | [`velesdb-core`](https://crates.io/crates/velesdb-core) (Rust) | The engine itself |\n| Give my agent memory | [`velesdb-memory`](crates/velesdb-memory) | MCP server + context compiler, any MCP client; `.mcpb` bundles on the [MCP Registry](https://registry.modelcontextprotocol.io/?q=velesdb-memory) |\n| Call it from Node | [`@wiscale/velesdb-memory-node`](https://www.npmjs.com/package/@wiscale/velesdb-memory-node) | Memory wedge ([full engine via server + TS SDK](crates/velesdb-node/README.md#need-the-full-engine)) |\n| Run it in a browser | [`@wiscale/velesdb-sdk`](https://www.npmjs.com/package/@wiscale/velesdb-sdk) | WASM, ~710 KB gzipped, fully client-side |\n| Serve it over HTTP | [`velesdb-server`](https://crates.io/crates/velesdb-server) | 54 REST endpoints — [API reference](docs/reference/api-reference.md) · [OpenAPI](docs/openapi.yaml) · [server security](docs/guides/SERVER_SECURITY.md) |\n| Ship on mobile/desktop | [`velesdb-mobile`](crates/velesdb-mobile) · [Tauri plugin](crates/tauri-plugin-velesdb) | iOS / Android / desktop |\n\nTool parity per surface is published honestly — including where a surface is still behind: [memory crate README](crates/velesdb-memory/README.md). Worked examples: [examples/](examples/README.md).\n\n
\nServing it over HTTP — the 54 REST endpoints, by category\n\n| Category | Key Endpoints |\n|----------|--------------|\n| **Collections** | `POST /collections`, `GET /collections`, `GET/DELETE /collections/{name}` |\n| **Points** | `/collections/{name}/points`, `/collections/{name}/points/raw`, `/collections/{name}/points/scroll`, `/collections/{name}/stream/insert`, `/collections/{name}/stream/enable`, `/collections/{name}/points/{id}/relations`, `/collections/{name}/points/{id}/ttl`, `/collections/{name}/relations` |\n| **Search** | `/collections/{name}/search`, `/collections/{name}/search/batch`, `/collections/{name}/search/hybrid`, `/collections/{name}/search/text`, `/collections/{name}/search/multi`, `/collections/{name}/search/ids`, `/collections/{name}/match` |\n| **Graph** | `/collections/{name}/graph/edges`, `/collections/{name}/graph/edges/{id}`, `/collections/{name}/graph/edges/count`, `/collections/{name}/graph/traverse`, `/collections/{name}/graph/traverse/stream`, `/collections/{name}/graph/traverse/parallel`, `/collections/{name}/graph/nodes`, `/collections/{name}/graph/nodes/{id}/degree`, `/collections/{name}/graph/nodes/{id}/edges`, `/collections/{name}/graph/nodes/{id}/payload`, `/collections/{name}/graph/search` |\n| **Indexes** | `GET/POST /collections/{name}/indexes`, `DELETE /collections/{name}/indexes/{label}/{property}`, `/collections/{name}/index/rebuild` |\n| **VelesQL** | `/query`, `/aggregate`, `/query/explain` |\n| **Admin** | `/health`, `/ready`, `/metrics`, `/guardrails`, `/collections/{name}/stats`, `/collections/{name}/config`, `/collections/{name}/flush`, `/collections/{name}/analyze`, `/collections/{name}/empty`, `/collections/{name}/sanity`, `/collections/{name}/compact`, `/collections/{name}/vacuum` |\n\n> **Full API reference:** [docs/reference/api-reference.md](docs/reference/api-reference.md) | **OpenAPI spec:** [docs/openapi.yaml](docs/openapi.yaml) | **Server security:** [docs/guides/SERVER_SECURITY.md](docs/guides/SERVER_SECURITY.md)\n\n
\n\n## How it compares\n\n| | **VelesDB** | Chroma | Qdrant | pgvector |\n|---|---|---|---|---|\n| **Architecture** | Vector + graph + columnar, unified | Vector only | Vector + payload | Vector extension for PostgreSQL |\n| **Metadata filtering** | Typed ColumnStore + secondary indexes | JSON scan | JSON payload | SQL |\n| **Graph support** | Native (`MATCH` clause) | No | No | No |\n| **Query language** | VelesQL (SQL + NEAR + MATCH) | Python API | JSON API / gRPC | SQL + operators |\n| **Embeddings from text** | Opt-in [local / OpenAI adapters](crates/velesdb-python/python/velesdb/embed.py); no bundled model | [Embedding functions](https://docs.trychroma.com/docs/embeddings/embedding-functions), with a default local model in Python/TypeScript | Opt-in client-side [FastEmbed](https://qdrant.tech/documentation/fastembed/fastembed-semantic-search/) | None; bring vectors from an external model |\n| **Deployment** | Embedded / Server / WASM / Mobile | Server (Python) | Server (Rust) | Requires PostgreSQL |\n| **Binary size** | ~14 MB | ~500 MB (with deps) | ~50 MB | N/A (PG extension) |\n| **Browser / Mobile** | Yes / Yes | No | No | No |\n| **Offline / Local-first** | Yes | Partial | No | No |\n\n> **Sweet spot:** vector + graph + structured filtering in one engine, local-first, auditable. **Not the best fit (yet):** a managed cloud service with a multi-node distributed cluster. Competitor figures are typical public ranges, not a head-to-head run we performed — [run your own](docs/BENCHMARKS.md). Detailed comparison against agent-memory products (Mem0, Zep, Letta), as of mid-2026: [docs/WHY_VELESDB.md](docs/WHY_VELESDB.md).\n\n## VelesDB Premium — the enterprise control plane\n\nThe core engine is source-available and stays that way. **Premium** adds the company-grade layer on top of the same binary, for organizations running agent fleets on sensitive data: **RBAC** on every endpoint including the memory and context-compiler surfaces · **audit trail** (who, what, when — metadata only, GDPR-conscious) with forensic replay · **multi-tenancy** with hard per-tenant isolation and two-level deletion rights · **clustering and air-gapped deployment** · a **WebAdmin** UI for operators.\n\nPricing on quote — **contact@wiscale.fr** · [velesdb.com](https://velesdb.com). Built by [Wiscale](https://wiscale.fr) (France; GDPR and data-sovereignty native).\n\n---\n\n\n\n## Known limitations — honest boundaries\n\nThe items below are deliberate trade-offs or Premium-tracked features, **not correctness gaps** — the Community Edition is production-ready for single-node, local-first deployments. We publish them next to the strengths, including the ones we have not fixed yet.\n\n| # | Limitation | Scope | Tracked |\n|---|---|---|---|\n| 1 | **Single writer per collection** — WAL is serialized; concurrent writers contend on the same fsync lock. | Design trade-off (local-first, crash-safe by default). Read throughput is unaffected. | Concurrent WAL writer planned for [Premium](#velesdb-premium--the-enterprise-control-plane). See [docs/CONCURRENCY_MODEL.md](docs/CONCURRENCY_MODEL.md). |\n| 2 | **No distributed replication** — single-node; no Raft, no sharding, no automatic failover in Core. | Deliberate: the sweet spot is local-first / embedded. | Raft-based replication tracked for Premium. |\n| 3 | **No advanced RBAC / multi-tenant isolation in Core** — Core ships the `DatabaseObserver` enforcement seam — live on every HTTP read path since 3.10.0 and still current in 4.0.0 — not the policy engine. | Core ships the hook, not the policy engine. | [Premium](#velesdb-premium--the-enterprise-control-plane) feature. |\n| 4 | **WASM MATCH limited to 2 hops** — 3+ hop `MATCH` works fully in native builds. | Browser-build scope limit, not a correctness issue. | Tracked. |\n| 5 | **SIFT1M fingerprint sidecar not yet committed** — the loader falls back to TOFU mode until the reference machine commits the pinned hashes. | Not a correctness issue — shape validation still applies. | Bootstrap shipped; sidecar pending. |\n| 6 | **No head-to-head Docker Compose benchmark vs Qdrant / Chroma / FAISS yet** — SIFT1M already gives literature-comparable numbers. | Side-by-side numbers need infrastructure not frozen yet. | Tracked. |\n| 7 | **Context-compiler tool parity varies by surface** — the MCP server and Rust have the full set; Node, Python and WASM are partially behind, and the WASM working contexts are intra-session only. | Binding scope, not an engine gap; MCP covers any client meanwhile. | Per-surface table in the [memory crate README](crates/velesdb-memory/README.md). |\n\nInternal technical limitations (query-planner approximations, plan-cache semantics): [docs/reference/KNOWN_LIMITATIONS.md](docs/reference/KNOWN_LIMITATIONS.md).\n\n## Contributing & contact\n\n**Quality bar:** `cargo test --workspace` — 9k+ tests across Rust, TypeScript and Python run in CI on every merge; exact commands in [QUALITY_BAR.md](QUALITY_BAR.md).\n\nContributions welcome — start with [CONTRIBUTING.md](CONTRIBUTING.md) and the [good first issues](https://github.com/cyberlife-coder/VelesDB/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22). Security reports: [SECURITY.md](SECURITY.md). Roadmap: [ROADMAP.md](ROADMAP.md) · [Changelog](CHANGELOG.md) · [DeepWiki](https://deepwiki.com/cyberlife-coder/VelesDB).\n\n**License:** [VelesDB Core License 1.0](LICENSE) (source-available). Premium: commercial license.\n**Contact:** contact@wiscale.fr · [velesdb.com](https://velesdb.com)\n\nThe name nods to Veles, a deity of old Slavic myth — a keeper of hidden knowledge and boundaries.\n", "bytes": 26360, "sha": "73bf934bf5cbf9a46cd47e128eae2fa38175f0adcaa40c08b2f89f5575ff8bb9", "repo_slug": "cyberlife-coder/velesdb", "fonte": "repo", "truncated": false, "api": "https://agentalog.com/api/listings/mcp_io_github_cyberlife_coder_velesdb_memory_22ab948d/readme" }