#search #encryption #knowledge-base #vault

app p3ak-vault

P3AK vault CLI — encrypted portable AI knowledge store. Create, ingest, search, and serve .vault files. 35 formats, hybrid BM25+vector search, 98% accuracy.

1 unstable release

0.1.0 Feb 27, 2026

#1057 in Cryptography

MIT license

565KB
12K SLoC

P3AK vault

Encrypted, portable, AI-searchable single-file knowledge store.

Part of the P3AK platform — use standalone or with P3AK room and P3AK harness.


What It Does

P3AK vault stores any document collection in a single encrypted .vault file with hybrid AI search built in. No database, no server, no cloud dependency. One file you can move, back up, or delete.

  • AES-256-GCM encryption at rest with Argon2id key derivation
  • Hybrid search — BM25 (Tantivy) + TF-IDF ZVec + PageIndex tree at ≥91% top-1 accuracy
  • 34 file formats — markdown, PDF, DOCX, HTML, CSV, images (OCR), audio (transcribe), and more
  • Hash-linked WAL — tamper-evident audit log of every read and write
  • Canary monitoring — automatic recall health checks
  • Multi-vault federation — search across multiple vaults in one query
  • .mdr support — ingests P3AK Documents (multi-layer versioned format)

Quick Start

# Install
cargo install p3ak-vault

# Create a vault
p3ak-vault create --path company.vault

# Ingest documents
p3ak-vault ingest --path company.vault --dir ./documents --recursive --room legal

# Search
p3ak-vault search --path company.vault --query "renewal terms" --mode hybrid

# REST API
p3ak-vault serve --path company.vault --port 8080

CLI Commands

Command Description
p3ak-vault create Create a new empty vault
p3ak-vault ingest Ingest a file or directory (34 formats)
p3ak-vault search Hybrid search with room/store filters
p3ak-vault read Read docs, goals, or WAL entries
p3ak-vault write Write goal, doc, plan, or review entries
p3ak-vault classify AI-classify documents into 10 knowledge stores
p3ak-vault delete Soft-delete a document
p3ak-vault export Export a store subset to a new vault
p3ak-vault sync Sync documents between two vaults
p3ak-vault watch Watch a directory and ingest new files
p3ak-vault canary-add Register a canary recall check
p3ak-vault canary-check Verify recall health (exit 2 if below threshold)
p3ak-vault accuracy-test Run benchmark against ground-truth queries
p3ak-vault serve Start REST API server (127.0.0.1 only)

Global flags: --path / $P3AK_VAULT_PATH, --passphrase / $P3AK_VAULT_PASSPHRASE


Supported Formats

Tier 1 (pure Rust): md, markdown, txt, text, log, csv, tsv, json, jsonl, svg
Tier 2 (compiled-in): pdf, html, htm, docx
Tier 3 (external tools): png, jpg, jpeg, webp, tiff, gif (OCR via tesseract), mp4, mov, webm, mp3, m4a, wav (transcribe via Groq/whisper), ai, eps (Ghostscript), epub, odt, rtf, pptx (pandoc)
P3AK native: .mdr (P3AK Document — extracts current layer for indexing)


SDKs

# Python subprocess SDK
from p3ak_vault import VaultClient
vault = VaultClient("/path/to/company.vault", passphrase="secret")
results = vault.search("renewal terms", mode="hybrid")

# Python native (PyO3 — sub-millisecond, no subprocess)
from p3ak_vault_native import P3AKVault
vault = P3AKVault.open("/path/to/company.vault", passphrase="secret")
results = vault.search("renewal terms", mode="hybrid")

Configuration

# ~/.p3ak/config.toml
[vault]
path = "/path/to/default.vault"

[security]
passphrase_env = "MY_VAULT_KEY"

Environment Variables

Variable Description
P3AK_VAULT_PATH Default vault path
P3AK_VAULT_PASSPHRASE Vault passphrase
ANTHROPIC_API_KEY Enables Claude Haiku LLM classifier fallback
GROQ_API_KEY Enables Groq Whisper audio transcription

Test Suite

cargo test --lib -p vault-core              # 265 unit tests
cargo test -p p3ak-vault-cli               # 54 integration tests
cargo test --test accuracy -p vault-core   # 12 accuracy tests
cargo test --test stress -p vault-core     # stress tests
cd packages/p3ak-vault-py && pytest        # 35 Python SDK tests
cd packages/p3ak-vault-native && maturin develop && pytest  # 44 native tests

Part of P3AK

Product Role
P3AK vault (this repo) Encrypted storage + search infrastructure
P3AK room AI data room + document management
P3AK harness Agent orchestration (FORGE protocol)

P3AK ecosystem docs


License

MIT — see LICENSE

Dependencies

~63MB
~1M SLoC