debgpt (0.7.6) unstable; urgency=medium
* Release 0.7.6.
+ Properly render the <think> blocks from DeepSeek models.
-- Mo Zhou <lumin@debian.org> Thu, 20 Mar 2025 11:26:52 -0400
debgpt (0.7.5) unstable; urgency=medium
* Release v0.7.5.
+ New frontends for llama.cpp, DeepSeek, Nvidia-NIM.
-- Mo Zhou <lumin@debian.org> Mon, 03 Feb 2025 23:32:54 -0500
debgpt (0.7.4) unstable; urgency=medium
* Build-Depends on python3-setuptools. (Closes: #1080577)
-- Mo Zhou <lumin@debian.org> Sun, 19 Jan 2025 13:53:24 -0500
debgpt (0.7.3) unstable; urgency=medium
* Release v0.7.3.
-- Mo Zhou <lumin@debian.org> Thu, 16 Jan 2025 20:03:13 -0500
debgpt (0.7.2) unstable; urgency=medium
* Release v0.7.2.
-- Mo Zhou <lumin@debian.org> Sat, 11 Jan 2025 09:31:27 -0500
debgpt (0.7) unstable; urgency=medium
* Release v0.7 (DebGPT has gone through major refactoring)
+ AI is heavily used to help the development through DebGPT itself
(including code generation, documentaiton generation, type annotations)
* Major changes (including big changes to CLI usage):
+ relicense from MIT to LGPL-3.0+
+ reorganize codebase to make it more readable and maintainable
* unified all readers into a unified read() function.
* overhaul mapreduce code to make it faster, more efficient and robust
+ implemented the VectorDB, the EmbeddingModel, and the VectorRetriever.
With those components, DebGPT will be able to do retrieval-augmented
generation (RAG). But that feature is not yet implemented in CLI.
+ Supported embedding frontends: OpenAI, Google
+ add xAI frontend support.
* For all changes please see CHANGELOG.md
-- Mo Zhou <lumin@debian.org> Sat, 09 Nov 2024 02:47:22 -0500
debgpt (0.6) UNRELEASED; urgency=medium
* This release is not uploaded to Debian.
* Major changes:
+ add inplace file editing support, with optional auto git commit
+ add a TUI configuration wizard for configurating debgpt
* Minor changes:
+ updates to code and doc. Add project art generated by AI.
-- Mo Zhou <lumin@debian.org> Sat, 09 Nov 2024 02:35:53 -0500
debgpt (0.5.93) unstable; urgency=medium
* v0.5.93 -- 2024-10-29
+ add google search support for context loading (--mapreduce|-x)
* change `:sbuild` into `sbuild:` for mapreduce
* change :policy and :devref into policy: and devref: for mapreduce
* cli: enable markdown rendering by default
* misc minor improvements.
* Overhaul README with new examples and usage instructions.
* v0.5.92 -- 2024-10-28
+ Add support for rendering LLM response markdown text with rich.
+ PDF file support is added for both `--pdf` and `--mapreduce` arguments.
* misc minor updates and bugfixes.
* v0.5.91 -- 2024-10-27
+ Add retry mechanism for rate limit exceeded errors.
+ mapreduce functionality now supports loading text from URLs.
-- Mo Zhou <lumin@debian.org> Tue, 29 Oct 2024 12:10:29 -0400
debgpt (0.5.90) unstable; urgency=medium
* Release v0.5.90
+ Introduce mapreduce functionality for any-length text processing.
See README for usage examples.
* frontend: add oneshot() method for all frontends
* README: introduce mapreduce usage example.
* use yapf instead of autopep8 for code formatting
* debgpt: remove --sbuild option
* cli: add short '-a' option for '--ask' argument in mapreduce configuration
* cli: add --amend option to git commit subcommand.
-- Mo Zhou <lumin@debian.org> Sun, 27 Oct 2024 01:03:14 -0400
debgpt (0.5.2) unstable; urgency=medium
* Release v0.5.2
* add support for Anthropic frontend integration
* add support for Google Gemini frontend integration
* cli: add support for loading latest sbuild build logs
* README: add interactive mode tips and commands
* cli: fix session index display in interactive mode prompt after /reset
-- Mo Zhou <lumin@debian.org> Sat, 26 Oct 2024 17:12:38 -0400
debgpt (0.5.1) unstable; urgency=medium
* Release v0.5.1
* frontend: reorder vLLM API key argument definition in CLI
* metadata: add PyPI classifiers to pyproject.toml
* frontend: bugfix: inherit OpenAIFrontend instead
* Apply autopep8 to all python files.
* cli: refactor config template generation with helper function
* cli: enhance fresh install detection with OpenAI base URL check
* `defaults: update OpenAI model to gpt-4o`
-- Mo Zhou <lumin@debian.org> Wed, 16 Oct 2024 12:14:12 -0400
debgpt (0.5.0) unstable; urgency=medium
* Release v0.5.0
* Now we have a reasonable set of supported frontends:
OpenAI, Ollama, Llamafile, vLLM, and ZMQ
(excluding the dryrun frontend for debugging)
* genconfig: fix toml grammar error.
* defaults: be verbose on loading config and overriding configs
* task: explicitly annotate debgpt usage for debgpt git commit
* frontend: add support for vLLM service integration
* frontend: use AbstractFrontend init directly instead of super()
-- Mo Zhou <lumin@debian.org> Wed, 16 Oct 2024 01:08:39 -0400
debgpt (0.4.95) unstable; urgency=medium
* Release v0.4.95.
* frontend: Skip console output during config template generation
* Add llamafile frontend (alias to openai frontend)
* Add ollama frontend (OpenAI-API compatibibility mode)
* Add --monochrome to disable colorized outputs during conversation
(Closes: #1061387)
* cli: Provide a fresh-install guide (Closes: #1064469)
-- Mo Zhou <lumin@debian.org> Wed, 16 Oct 2024 00:25:38 -0400
debgpt (0.4.94) unstable; urgency=medium
* Release v0.4.94. (Closes: #1060654)
Find the full changelog at /usr/share/doc/debgpt/changelog-full.md.gz
-- Mo Zhou <lumin@debian.org> Fri, 12 Jan 2024 21:35:27 -0500
debgpt (0.4.93) unstable; urgency=medium
* Release v0.4.93.
-- Mo Zhou <lumin@debian.org> Tue, 09 Jan 2024 12:00:36 -0500
debgpt (0.4.92) unstable; urgency=medium
* Release v0.4.92.
-- Mo Zhou <lumin@debian.org> Mon, 08 Jan 2024 23:39:14 -0500
debgpt (0.4.90) unstable; urgency=medium
* Initial release. (Closes: #1060113)
-- Mo Zhou <lumin@debian.org> Sun, 07 Jan 2024 12:09:53 -0500