LLM 0.32a1 Released: Fixes Tool-Calling Conversation Storage Bug
LLM 0.32a1 fixes a bug preventing tool-calling conversations from reloading correctly from SQLite.
Simon Willison released LLM 0.32a1, an alpha bugfix addressing a critical issue from 0.32a0 where conversations involving tool-calling could not be properly reinflated from the SQLite database. The fix ensures tool-calling conversation histories can be correctly restored and continued. Users affected by this issue are encouraged to upgrade immediately.
Overview
Simon Willison has released version 0.32a1 of his open-source command-line tool LLM. This is an alpha bugfix release that primarily addresses a critical bug from the previous version (0.32a0) where tool-calling conversations failed to reload correctly from the SQLite database.
What Was Fixed
SQLite Storage Issue with Tool-Calling Conversations
This update fixes a bug introduced in version 0.32a0: when users had conversations involving tool-calling, those conversation records could not be properly reinflated when read back from the SQLite database. This meant that previously saved conversation histories containing tool calls could not be properly restored or continued.
The issue was tracked in GitHub Issue #1426 and has been resolved in this release.
Background: Tool-Calling in LLM
LLM is a powerful command-line tool developed by Simon Willison that allows users to interact with various large language models through the terminal. The tool-calling (function-calling) capability added in recent versions is one of the core features of modern LLM applications, enabling models to invoke external tools and functions to complete tasks during conversations.
Persistent conversation storage in SQLite is an important feature of the LLM tool, allowing users to review past conversations and continue previous interactions. Correct serialization and deserialization of tool-calling conversations is critical for maintaining the integrity of conversation context.
Version Notes
It's worth noting that 0.32a1 is still an alpha version (indicated by the "a" suffix), meaning it's in an early testing phase. For production use, it's recommended to watch for subsequent stable releases. However, for users already running 0.32a0 who have encountered the tool-calling conversation storage issue, upgrading to this version is recommended as soon as possible.
How to Update
Users can update to the latest alpha version with the following command:
pip install llm==0.32a1
Or to install the latest version including pre-releases:
pip install --pre llm
Related articles
Tech FrontiersBuoyant Open-Sources Kinference: An Inference Control Plane for vLLM on K8s
Buoyant open-sources Kinference, an inference control plane for vLLM on Kubernetes, with declarative routing, KV-cache-aware load balancing, and per-tenant metrics—claiming 95% lower median TTFT in official benchmarks.
Tech FrontiersThe Truth Behind Anthropic's 2x R&D Speedup: A Deep Dive into the METR Report
METR's external evaluation shows Anthropic's AI testing progress over the past year roughly equals the prior two years, with R&D accelerating about 2x—and roughly 80% likely from AI assistance. The report also admits key evidence was redacted, the range spans 1.3x to 3x, and Anthropic says it hasn't doubled yet, diverging from METR's judgment.
Tech FrontiersT3 Code: Free, Open-Source, and a Unified Way to Manage All AI Coding Agents
T3 Code is a free open-source app that aggregates AI coding tools like Cloud Code and Codex into one interface, with conversation handoff, branch isolation review, automatic PR creation, and remote access.