What's Changed

• Use RList instead of hardcoded NEWS vectors by @trufae in #215
• Fix editing r2js scripts in auto mode
• Handle comments in commands in all the auto calls
• Add r2ai.debug to see the llm requests and responses
• Prompt before leaving the interrupted auto mode
• Comments in commands are now used as annotations for the conversation
• Increase timeout to 4 minutes
• Resizing the terminal is no longer interrupting the network calls
• Implement r2ai -C to compact the message log
• Add r2ai.auto.think to disable reasoning
• Implement the -LR command to create a log report
• Capture the thinking messages
• Oh great now we have a bug in the json serializer of r2
• Add r2ai -b for the shell to change the base address
• Use R2_CURL on Windows (may work by default if curl is in PATH)
• Implement r2ai -E and re-run the script
• Proper command escapes for curl and pwsh http backends

Full Changelog: 1.2.4...1.2.6

What's Changed

• Th by @trufae in #214

Full Changelog: 1.2.2...1.2.4

What's Changed

• Fix decai catch handling issue
• Add manpage
• Add -E to edit ~/.config/r2ai/rc file
• Support vscode-style prompt files (yaml+markdown)
• Use jan-nano by default (2GB model)
• Simplify sleep_retry logic using r_sys_sleep for portability
• Implement r2ai.auto.raw to use tool calling with non-tools models by @trufae in #213

Full Changelog: 1.2.0...1.2.2

Main Changes

• Use ollama/gpt-oss:20b as default provider/model
• Refactor many hardcoded actions into the new query prompts
• Add -q to query user defined prompts (see the ./prompts directory)
• Use powershell as fallback when curl is not available on windows
• Remove support for old versions of radare2 (6.0 at least)

What's Changed

• fix tmp dir on android(termux) by @AbhiTheModder in #208
• Fallback to powershell when curl is not found on Windows by @trufae in #209
• Use clang-format-radare2 instead of the hacky indent.py by @trufae in #210
• Noglob by @trufae in #211
• Prompts by @trufae in #212

New Contributors

• @AbhiTheModder made their first contribution in #208

Full Changelog: 1.1.2...1.2.0

What's Changed

• more references by @cryptax in #204
• Include the r2ai plugin in the meson ci builds by @trufae in #206

Full Changelog: 1.1.1...1.1.2

Full Changelog: 1.1.0...1.1.1

What's Changed

• Fix listing ollama models by @cryptax #193
• Removing port in baseurl for r2ai C by @cryptax in #195
• Support lm-studio remote servers by @cryptax in #196
• Support setting api keys via environment variable
• Fix support for gpt-5
• Use r2-6.0.2 by @trufae in #197
• Add the Windows builds in the CI by @trufae in #202
• Refactored decai.r2.js by @kxynos in #203

New Contributors

• @kxynos made their first contribution in #203

Full Changelog: 1.0.0...1.1.0

What's Changed

• Support r2 6.0
• Added new logo (skynet like)
• Unify and rename "ask_to_execute" to "yolo" everywhere
• Add -e model=? to list models by @dnakov in #179
• remove static lists for providers where dynamic fetch works fine by @cryptax in #180
• Dockerfile for r2 + r2ai + decai + r2ghidra + nice .radare2rc by @sha0coder in #181
• Adding max_input_tokens to auto C by @cryptax in #184
• Use "baseurl" instead of "host"/"port" configurations for all api providers
• Support LMStudio
• Add http.files to handle large data thru curl (larger than commandline arguments support)
• Add TTS support to Decai
• Support Gemini
• Support r2pipe2 calls to capture error messages and logs
• Add docker image to ease setup
• Improved auto repl in r2ai and decai
• Support larger contexts (>128k in local and much more with remote ones)
• List available models in r2ai like it did in decai

New Contributors

• @sha0coder
• @cryptax

Full Changelog: 0.9.6...1.0.0

What's Changed

• Fix non-auto opts like -x -d etc by @dnakov in #157
• Add support for reasoning_content by @dnakov in #158
• R2aic model names by @cryptax in #171
• clean model name by @emilius3m in #169
• Add system install for r2ai-plugin by @prodrigestivill in #174

New Contributors

• @emilius3m made their first contribution in #169

Full Changelog: 0.9.4...0.9.6

What's Changed

• Fix #128 by @dnakov in #129
• Doc r2ai by @cryptax in #127
• Decai auto by @cryptax in #138
• fix: empty command in auto mode with ask_to_execute enabled by @stevenengland in #135
• [WIP] Auto mode, rewrite everything (r2ai.c) by @dnakov in #141
• Add .clang-format; print tool calls and responses (r2ai.c) by @dnakov in #142
• Add stats display; Support for all the other providers by @dnakov in #143
• Implement chat log; Add tokens to stats by @dnakov in #144
• Use libcurl if available by @dnakov in #145
• Use R2_CURL; Add javascript tool; Add Markdown renderer by @dnakov in #146
• Add http retry with expo backoff by @dnakov in #147
• Add ask_to_execute support; Add mistral and grok as providers by @dnakov in #148
• Error handling with retry; Fix some known differences between provide… by @dnakov in #149
• Add support for anthropic thinking by @dnakov in #152
• Remove adding additional empty user message by @dnakov in #153

New Contributors

• @stevenengland made their first contribution in #135

Full Changelog: 0.9.2...0.9.4