llm_ollama 0.1.9

Ollama backend implementation for LLM interactions. Provides streaming chat, embeddings, tool calling, vision support, and model management.

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

0.1.9 - 2026-02-28

Added

• Integration regression coverage for stream boundary resilience in tool-call loops (read_file -> write_file -> final assistant response)

Changed

• OllamaStreamConverter.toLLMStream() now uses boundary-safe NDJSON line framing across transport chunks
• Stream parsing now uses bounded malformed-line retries (3 consecutive malformed non-empty lines) and throws explicit LLMApiException after budget exhaustion instead of silently dropping indefinitely
• maxToolAttempts default increased from 25 to 90
• Bumped llm_core dependency to ^0.1.9

0.1.8 - 2026-02-26

Added

• OllamaMessageConverter.messagesToOllamaJson() for list-aware conversion; derives tool_name from toolCallId via preceding assistant's tool_calls or synthetic ID parsing
• Fallback chain for tool_name: lookup by id, parse synthetic tool_N_name, or send tool_call_id only (Ollama supports both)
• Thorough tool response integration tests (15 cases) validating stream contract, deterministic results, error handling, tool chains, and Ollama-specific behavior

Changed

• Tool messages now converted via messagesToOllamaJson(); tool_name encapsulated in Ollama layer, derived from toolCallId
• Replaced per-message toJson() with list-aware messagesToOllamaJson() for correct tool message conversion

0.1.7 - 2026-02-10

Added

• batchEmbed() implementation: delegates to existing batch-capable embed() (Ollama /api/embed accepts an array of inputs).

0.1.6 - 2026-02-10

Fixed

• Ensured Ollama tool calls always produce LLMToolCall instances with non-null, non-empty id values, synthesizing IDs when Ollama does not provide them.
• Aligned tool-calling behavior with llm_core's toolCallId validation so that tool execution no longer fails with Tool message must have toolCallId when used together with llm_core.

0.1.5 - 2026-01-26

Added

• Builder pattern for OllamaChatRepository via OllamaChatRepositoryBuilder for complex configurations
• Support for StreamChatOptions in streamChat() method
• Support for chatResponse() method for non-streaming complete responses
• Support for RetryConfig and TimeoutConfig for advanced request configuration
• Input validation for model names and messages

Changed

• streamChat() now accepts optional StreamChatOptions parameter
• Improved error handling and retry logic
• Enhanced documentation

0.1.0 - 2026-01-19

Added

• Initial release
• Ollama backend implementation for LLM interactions:
  • Streaming chat responses
  • Tool/function calling support
  • Vision (image) support
  • Embeddings
  • Thinking mode support
  • Model management (list, pull, show, version)
• Full compatibility with Ollama API