chonkie Releases

A feature-heavy release expanding international PII coverage with recognizers for German, Swedish, and Canadian identities, plus a new 'Slim NLP Engine' for performance and CI/CD improvements.

Jun 28, 2026

A performance and security-focused release that includes UI optimizations, the removal of unused Docker dependencies, and a major TypeScript upgrade.

Jun 24, 2026

A major feature release introducing the `triage_cve` orchestrator for concurrent security analysis, expanding data sources, and implementing significant security hardening measures.

Jun 22, 2026

This release establishes a global data layer with 18 sources and 18 read-only tools, expands the broker connector roster to 10, and introduces the Alpha compare full stack. It also overhauls provider reliability and launches Research Autopilot Phase 1.

Jun 19, 2026

This release adds concurrent session support via a pipeline pool, introduces speculative turn revisions for better performance, and improves handling for Paraformer and speculative VAD.

Jun 11, 2026

ECC 2.0.0 establishes ECC as a cross-harness operating system, featuring a local Control Pane, MCP Inventory, and 261 public skills. It marks the stable graduation of the 2.0 line.

Jun 10, 2026

This release adds PyEmscripten wheels for web compatibility, fixes non-ASCII text handling in JSON exports, and supports Gemini embedding dimensions in Catsu wrappers.

Jun 1, 2026

v1.6.7 introduces MistralOCR integration, allowing users to extract text from images directly within Chonkie.

May 19, 2026

A significant performance update to the CodeChunker, reducing indexing time by 44.5% and improving search quality metrics.

May 13, 2026

Enhances evaluation and CLI tooling by persisting performance metrics and adding specific evaluation pipelines for frozen pipecat models.

May 10, 2026

This release introduces support for agentic skills via skills.sh and replaces all tokenizer backends with the 'tokie' library.

May 6, 2026

Introduces mining features for data uploads and updates model versions to support Chinese language processing.

May 4, 2026

Adds a notebook for loading smart turn models and establishes a complete training and evaluation pipeline for Chinese smart turn models.

Apr 30, 2026

v1.6.4 fixes a critical bug where importing Chonkie failed due to an unconditional pandas import, resolving the issue for users not installing the `[table]` extra.

Apr 21, 2026

v1.6.3 introduces a LanceDB handshake mechanism, adds filename metadata, improves Markdown support, and adds a table-to-JSON converter.

Apr 20, 2026

Initial release of the CVE MCP Server, establishing the foundational tools and MCP server setup.

Apr 14, 2026

This release adds the TeraflopAI Chunker, allowing users to leverage external AI for chunking via API, alongside various dependency updates and bug fixes.

Apr 7, 2026

ECC v1.10.0 focuses on harness reliability, cross-platform compatibility, and better operator workflows, serving as a precursor to the upcoming ECC 2.0 Alpha release.

Apr 5, 2026

This patch release focuses on fixing import issues, updating dependencies (like pyjwt and orjson), and adding httpx support.

Mar 18, 2026

Enhances model inference with GPU support and ONNX Runtime integration. Significantly expands PII coverage with recognizers for UK, Nigeria, and US NPI, alongside configurable regex timeouts and security hardening.

Mar 18, 2026