v0.88.37

What's Changed

MR: #216
Full Changelog: v0.88.36...v0.88.37

Includes:

• Centralizing model response chunk management to better deal with model output format variations

v0.88.36

What's Changed

MR: #212
Full Changelog: v0.88.35...v0.88.36

Includes:

• Command aliases
• Less noisy debug tracer
• Update model metadata

v0.88.35

What's Changed

MR: #210
Full Changelog: v0.88.34...v0.88.35

Addresses:
#209

v0.88.34

What's Changed

MR: #207
Full Changelog: v0.88.33...v0.88.34

Addresses:
#202
#205

Includes:

• Exclude images from scraping outputs since they blow up the context window
• More graceful exit logic
• Increase agent mode message cache efficiency by ~5% by adding a bit of adaptivity to the main context chunking
• For the tweak logic
  • Aider-AI#3085

v0.88.33

What's Changed

MR: #200
Full Changelog: v0.88.32...v0.88.33

Addresses:
#33
#92
#141
#197

Includes:

• Fixes output escaping to account for the fact that rich translates bracket enclosed strings as style tags

v0.88.32

What's Changed

MR: #199
Full Changelog: v0.88.31...v0.88.32

Includes:
#198 - Fix for linear output mode with new chat_chunks.py configuration

v0.88.31

What's Changed

MR: #195
Full Changelog: v0.88.30...v0.88.31

Includes:

• #194 - Gemini Pro 2.5 Fixes
• #196 - Proper version update messaging text
• .gitattributes to enforce lf line endings on commit
• Improvements to agent mode task time horizon and cache-ability (roughly 60-70%)
  • Reorder chunks for agent mode
  • Split current messages into old and young sub sections
  • Sort file contents by last edit time of file
  • Add old messages before file edits and young messages after editable files

New Contributors

• @szmania made their first contribution in #194

v0.88.30

What's Changed

MR: #191
Full Changelog: v0.88.29...v0.88.30

Includes:

• Better caching heuristic, sorted as system prompt, read only files, writeable files, repo map, done message, current messages
• Add model partial outputs to bottom toolbar to give a little indication of what the model is processing while generating tool calls
• Fix auto save

v0.88.29

What's Changed

MR: #190
Full Changelog: v0.88.28...v0.88.29

Addresses:
#187
#189

Includes:

• Version updates in requirements.txt
• #186: Repo Map profiling for getting information necessary for further optimizations
• #188: Preserving new lines in mixed line ending repos
• Allows raw JSON responses in agent mode to be interpreted as tool calls for weaker models
• Add a "Thinking" tool that's really a place for a model to put free form thought and notes, ideally to keep itself on track during generation

New Contributors

• @pjcreath made their first contribution in #188

v0.88.28

What's Changed

MR: #185
Full Changelog: v0.88.27...v0.88.28

Addresses:
#184

Includes:

• Fix hang from issue 184 (above) by making sure output stream is None'd on finish, even if the generation task errors
• Refactor local tools to implement a common BaseTool abstract class as a precursor to a plugin system
• ReplaceText tool to display a diff of its output so it's more legible what is changing during generation
• Update with Aider main