Changelog¶
All notable changes to Ask Chitrank.
[0.7.0] - 2026-03-22¶
Features¶
- Upgrade Groq LLM from Llama 3.3 to Llama 4 Scout 17B-16E across configuration and documentation.
- Add local query expansion and embedding provider fallback. Add synonym-based query expansion to improve retrieval recall for short/ambiguous queries without extra API calls. Add local embedding fallback (all-MiniLM-L6-v2) via EMBEDDING_PROVIDER config so the app degrades gracefully if Voyage AI's free tier changes.
- Upgrade LLM to Llama 4 Scout, add local embedding fallback, and implement synonym-based query expansion.
Maintenance¶
- Bump askchitrank package version to 0.6.3 in uv.lock.
[0.6.3] - 2026-03-22¶
Documentation¶
- Document PaaS ipv4 pooler requirements and exact-match cache architecture
- Clarify top_score mechanics in retrieval layer
- Expand detailed Mkdocs documentation regarding exact match cache & top_score
Features¶
- Implement exact match cache lookup in the chat pipeline to serve identical questions directly from cache and add traceback to error logging.
- Implement advanced retrieval scoring with lexical and source boosts, update architecture diagrams, and add PaaS deployment and connection pooling documentation.
Maintenance¶
- 🔒 pypdf has inefficient decoding of array-based streams
- Decrease retrieval minimum similarity threshold.
[0.6.2] - 2026-03-21¶
Documentation¶
- Updated documentation
Features¶
- Update bruno api documentation
- Added testing for BL
- Introduce dev mode for local development and enhance chat safety with new routing and prompt hardening.
- Update README with Vite chat widget details and enhance chat safety tests to include latency verification in done events.
Maintenance¶
- Added seeded data for testing mode under /src/dev
[0.6.1] - 2026-03-18¶
Documentation¶
- Update project version to 0.6.1 and add live API link
[0.6.0] - 2026-03-18¶
Documentation¶
- Add descriptive comments to Makefile targets and commands.
Features¶
- Implement unified API token authentication and custom error handling across API endpoints, and refine the ingestion pipeline.
Maintenance¶
- Reorder MkDocs navigation entries.
Refactoring¶
- Restructure Bruno API collection into granular tests and enhance API documentation with auth, validation, and error details.
[0.5.0] - 2026-03-18¶
Documentation¶
- Add API and chat layer overviews, update project progress, and rename Sanity webhook secret configuration.
Features¶
- Implement source-diverse retrieval by limiting results per source and refine ingestion to categorize testimonials separately.
- Implement RAG chat functionality with Groq LLM integration and streaming capabilities.
- Introduce a new FastAPI server with rate limiting, accessible via a dedicated CLI command.
- Establish core API with v1 endpoints for health, streaming chat, and Sanity content ingestion, including their respective schemas and routing.
- Add
streamoption to chat requests, allowing clients to choose between SSE streaming and a single JSON response. - Introduce Bruno API collection with local environment, chat, health, and ingest endpoints.
- Implement API security headers, update documentation with new concepts and Sanity webhook setup, and add cache invalidation to the ingestion pipeline.
Maintenance¶
- Update project name from Folio AI to Ask Chitrank in changelog and config.
- Update GROQ_MODEL to llama-3.3-70b-versatile in configuration and documentation.
[0.4.5] - 2026-03-18¶
Bug Fixes¶
- Corrected author's last name in mkdocs.yml.
[0.4.4] - 2026-03-18¶
Bug Fixes¶
- Dead retrieval docs
[0.4.2] - 2026-03-18¶
Documentation¶
- Mark Phase 3 of the retrieval layer as complete in the project roadmap.
Features¶
- Implement semantic response caching and knowledge base similarity search.
[0.4.1] - 2026-03-18¶
Refactoring¶
- Use
Path.openand remove unused website URL parsing logic in the LinkedIn loader.
[0.4.0] - 2026-03-18¶
Documentation¶
- CHANGELOG project name references
- Expand tech stack documentation with new sections and tools, and refine database schema descriptions.
Features¶
- Implement a document ingestion pipeline to load, chunk, and embed data from PDF and Sanity CMS.
- Implement URL-based resume ingestion and Sanity API token authentication, with improved logging and documentation.
- Update build configuration and menu script for improved functionality.
- Add 'linkedin' as an ingestion source, remove 'all' option, and update default ingestion sources.
- Add LinkedIn profile data ingestion from CSV exports.
- Add functionality to load and extract text from local PDF files and update
greenletdependency. - Revamp LinkedIn data ingestion to use Profile.csv and filter Recommendations_Received.csv, replacing previous CSVs and adding profile formatting.
- Implement section-aware chunking for resumes with fallback to word-count chunking and update data path retrieval.
- Update resume ingestion to use local PDF, switch LinkedIn ingestion to CSVs, and refactor PDF text extraction
- Expand knowledge base to include LinkedIn, update project roadmap, and refine documentation.
- Add ingestion and retrieval overview documentation and update mkdocs navigation.
Maintenance¶
- Bump project version to 0.4.0.
Refactoring¶
- Use
pathlib.Path.openinlinkedin_loaderand addstrict=Falsetozipinpipeline.
[0.3.0] - 2026-03-17¶
Documentation¶
- Update project name references in CONTRIBUTING.md from ML-Notebook-Library to askchitrank.
- Add dedicated tech stack documentation and enhance project overview in the README and main documentation.
Features¶
- Introduce a new database module with SQLAlchemy async connection management and ORM models for knowledge chunks, response cache, and conversations.
- Initialize Alembic for database migrations and define the initial schema for conversations, knowledge chunks, and response cache.up
- Set up MkDocs and add initial documentation for development setup, architecture, and database.
- Add or update Makefile target to execute the
menu.pyscript. - Update Makefile to include new build targets and dependencies.
Refactoring¶
- Adjust API lifespan management and event handling.
[0.2.0] - 2026-03-17¶
Documentation¶
- Add CONTRIBUTING.md and CODE_OF_CONDUCT.md to establish contribution guidelines and community standards.
Features¶
- Establish core configuration, logging, and warning suppression utilities.
- Rename project to
folio-aitoaskchitrankand
Refactoring¶
- Adjust API lifecycle management.
[0.1.0] - 2026-03-17¶
Features¶
- Initialize project with core structure, dependency management, development tools, and GitHub templates.