Changelog

​

File upload (path-based)

• File upload - async ingest with chunking, batch summaries (every 100 chunks), and embedding
• List files, get file by id, delete file index
• File job status - poll upload and delete jobs (file_ingest or file_delete)
• Fetch text data - cursor pagination, up to 100 items per page

• upload-file, list-files, file-get, delete-file, file-job-status, fetch-text-data

• client.files.upload, list, get, delete, job_status
• client.documents.fetch_text_data for paginated text chunks
• MoorchehClient resource layout (namespaces, documents, files, similarity_search, answer) - MoorchehApiClient kept as legacy wrapper

• namespace on each result; summary_text on top hit via bare summary_chunk_id
• #key:value metadata filters and #keyword text filters in the query string (cloud parity)

• file_registry.json under ~/.moorcheh/data

​

AI Answer generation

​

What’s new

• Answer endpoint - Search Mode (namespace + RAG) and Direct AI Mode (namespace: "")
• LLM configuration saved under llm in ~/.moorcheh/config.json; prompted during moorcheh configure
• moorcheh answer CLI command and client.answer.generate()
• Health and status include llm_provider and llm_model
• Structured output - optional structured_response: { "enabled": true } with default JSON schema
• Docs: Generate AI Answer, Python client, CLI

​

Supported LLM models (examples)

​

Upgrade notes

• Re-run moorcheh configure --force to set LLM provider and model, or add a llm block to config.json manually.
• Rebuild or pull an updated moorcheh/server image that includes the answer endpoint.

​

Multi-provider text embeddings

​

Install and upgrade

pip install moorcheh-client==0.1.2
# or
pip install -U moorcheh-client

​

What’s new

• moorcheh configure - Interactive setup for provider, model, and API key (cloud providers)
• moorcheh up - Reads saved config; runs the embedding wizard on first start if config is missing
• Ollama (optional) - Bundled moorcheh-ollama starts only when provider is ollama and host Ollama is not running; embedding model is pulled automatically when missing
• OpenAI / Cohere - Server container only; no Ollama required; API keys stored locally in config.json
• Health - GET /health and moorcheh status include embedding_provider and model
• CLI flags - --embedding-provider, --embedding-model, --embedding-api-key, --no-configure, --skip-ollama-model-pull
• Docs - Embedding providers, moorcheh configure, updated Prerequisites and moorcheh up
• Tests - Coverage for config, Ollama setup, Docker runtime, and CLI embedding flows

​

Supported embedding models

​

Upgrade notes

• Run moorcheh configure once after upgrading the client (or answer prompts on the next moorcheh up).
• Existing data under ~/.moorcheh/data is unchanged.
• If you previously relied on Ollama-only defaults, pick ollama in configure to keep the same behavior, or switch to a cloud provider.

​

Client improvements

​

Install and upgrade

pip install moorcheh-client==0.1.1
# or
pip install -U moorcheh-client

​

What’s new

• Python 3.10–3.13 - client package tested across supported Python versions
• Quality - expanded automated tests for the API client, CLI, and Docker runtime helpers
• Packaging - more consistent PyPI releases for moorcheh-client

​

Upgrade notes

​

Initial release

​

Runtime

• moorcheh up / down / status - Start and stop the Moorcheh API server (moorcheh/server:latest) via Docker Compose
• Ollama integration - Host Ollama or bundled moorcheh-ollama container for embeddings (default model: nomic-embed-text)
• Local data - Persistent storage under ~/.moorcheh/data

​

REST API

• Health - Server status, embedding model, and global item quota (GET /health)
• Namespaces - Create, list, and async delete text or vector namespaces
• Data - Async upload of documents (text) and precomputed vectors, with job polling
• Items - Get and delete items by id (up to 100 ids per request)
• Search - Semantic search with text or vector queries across one or more namespaces

​

CLI and Python client

• moorcheh-client on PyPI - Installs the moorcheh CLI and MoorchehApiClient in one package
• Full parity - CLI and Python SDK cover the same API surface (namespaces, uploads, items, search)

​

Limits

• Unlimited namespaces - no cloud-tier namespace count limit; only the global item cap applies
• 100,000 items global cap across all namespaces
• Item ids unique per namespace