Skip to content

v0.5.1

Choose a tag to compare

View all tags
@github-actions github-actions released this 18 Oct 13:16
· 4 commits to main since this release
v0.5.1
0e6490c
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Changed

  • README rewritten to match v0.5.x: what inode does, install requirements per platform (Go 1.26, a C compiler for SQLite on every platform), quick start on the Ollama path, per-backend defaults, MCP setup for Claude Code and Cursor, every command, and links to the docs instead of an outdated roadmap table. inode --help describes the tool the same way. (#80)
  • The default model for llm.backend = claude-api is now claude-sonnet-5-5 (was Claude Sonnet 4.5, which reaches end of life on November 30, 2026). Only applies when llm.model is not set. Dependencies updated: Anthropic SDK v1.81.0, pgx v5.11.0, pgvector-go v0.4.1, go-sqlite3 v1.14.52, golang.org/x/crypto v0.58.0. (#113)

Fixed

  • Choosing llm.backend = openai or claude-api, or embedding.backend = voyage, now uses that backend's own defaults. Before, every backend inherited Ollama's: OpenAI requests went to http://localhost:11434 with model llama3.2, Claude was asked for llama3.2, and Voyage got nomic-embed-text and a 768-dimension vector table instead of voyage-3 and 1024. Defaults per backend: Ollama llama3.2 / nomic-embed-text (768) at http://localhost:11434, OpenAI gpt-4o-mini at https://api.openai.com/v1, Claude claude-sonnet-5-5, Voyage voyage-3 (1024). Values you set still win, and inode config show prints the values in use, marking defaults as (default for <backend>). The v0.5.0 workaround of setting llm.model and llm.base_url for OpenAI is no longer needed. Change: an embedding model other than nomic-embed-text or voyage-3 now needs embedding.dimension set; inode stops with a message saying so instead of assuming 768. (#31)
  • inode config set accepts every setting inode reads. db.backend, db.dsn, search.top_k, search.max_distance and mcp.reveal_sensitive were rejected as unknown keys although the README told you to set them. Values are now checked (search.top_k abc or db.backend mysql fail with a message naming the expected type or values) and written with their type, so top_k = 10 instead of top_k = "10"; files written by earlier versions still load. inode config set --help lists every key, and inode config show prints every setting with API keys and db.dsn redacted. (#81)
  • Short note IDs no longer pick a note at random. inode note get, inode note delete and MCP get_note used to take the first note whose ID started with the given text, so inode note delete 3 could delete any note starting with 3, and an empty ID or % matched any note. A prefix that matches more than one note is now refused with the list of matches, IDs may only contain hex digits and -, note delete needs at least 4 characters and its confirmation names the note (short ID, summary, category) instead of repeating your input. (#90)

Full Changelog: https://github.com/BiSemaphore/inode/blob/main/CHANGELOG.md

Assets 4
Loading