V0.1.6

[krazyjakee]

Key Features & Enhancements

• Pluggable Embedding Backends
  Integrates a new agentsdb-embeddings crate, enabling support for a wide array of embedding providers including:

  • Local: Candle, ORT/FastEmbed
  • Remote: OpenAI, Voyage, Cohere, Anthropic, Bedrock, Gemini

• Comprehensive Options Management
  Introduces the agentsdb options command for flexible configuration of embedding settings, including:

  • Backend selection
  • Model choice
  • Embedding dimensions
  • Caching
  • API base URLs
  • API key environment variables
    Features include an interactive setup (wizard) and a checksum allowlist for local models.

• Robust Import/Export Functionality
  Adds agentsdb export and agentsdb import commands, facilitating the exchange of layer data in stable JSON/NDJSON formats, with support for:

  • Redaction
  • Deduplication

• Accelerated Search with Indexing
  Implements an agentsdb index command to build rebuildable sidecar indexes, significantly improving the performance of exact search queries.

• Streamlined Proposal Workflow
  Introduces agentsdb proposals commands (list, show, accept, reject) for managing MCP promotion requests.
  Proposals are now durably stored as meta.proposal_event chunks directly within AGENTS.delta.db, eliminating reliance on external sidecar files.

• Enhanced Diffing Capabilities
  The agentsdb diff command now includes a --target user option to identify potential conflicts when promoting changes to the user layer.

• Modernized Web UI
  The agentsdb web interface has been completely revamped, offering:

  • Intuitive browsing of layers and chunks
  • Proposal management
  • Import/export operations
    Additional features include Markdown rendering for chunk content and a theme toggle.

---

Core Changes & Improvements

• AGENTS.db Format Update
  The underlying file format now includes a LayerMetadata section, storing crucial embedding profile and tool information. This enables robust schema compatibility checks across layers.

• Unified Embedding Integration
  Core compile and write commands now leverage the new agentsdb-embeddings crate, ensuring consistent embedding generation based on rolled-up configuration.
  The --dim argument is now optional, with dimensions inferred from global or layer-specific options.

• Improved Atomic File Operations
  Updates to agentsdb-format ensure that write_layer_atomic and append_layer_atomic correctly handle and preserve layer metadata during file writes.

• Strengthened Error Handling
  Enhanced permission checks and clearer error messages are now provided for various file operations, particularly around writable layers.

• Extensive Dependency Updates
  Cargo.lock and Cargo.toml have been significantly updated to incorporate new crates and dependencies, ensuring a stable build environment.

---

Build & Documentation

• Updated CI/CD Workflows
  GitHub Actions (release.yml) and local installation scripts (install-local.sh) now include frontend build steps and enable all embedding features by default.

• Comprehensive Documentation
  The following documents have been updated to reflect new features and architectural changes:

  • README.md
  • RFC.md
  • Reference Implementation.md
    New dedicated documentation is also available:
  • Export.md
  • embedding.md

• Code Quality Standards
  A new .clippy.toml file enforces stricter linting rules, contributing to improved overall code quality.

• Expanded Environment Variable Examples
  .env.example has been updated to include environment variables for all supported embedding providers.