* fix(cli): expand model capability detection to include Llama, Nemotron, and Mistral models
The isModelCapable function was showing false warnings for Llama, Nemotron,
and Mistral models, claiming they had "limited reasoning and tool calling
capabilities" when they actually have excellent capabilities.
**Changes:**
- Added /llama/, /nemotron/, /mistral/ patterns to capability detection regex
- Updated tests to reflect that these model families ARE capable
- All tests passing (26/26)
**Research validation:**
- Llama 3.3/Nemotron: #1 on alignment benchmarks, Arena Hard 85.0
- Mistral: 81.2% MMLU, supports function calling and JSON mode
- Both families widely used for agent workflows with proven tool calling
**Impact:**
- Removes false warnings for users of these popular model families
- Enables proper multiEdit tool usage for capable models
- Aligns detection with real-world model capabilities
Tested with nvidia/Llama-3_3-Nemotron-Super-49B-v1 on MITRE AIP endpoints.
Authored by: Aaron Lippold <lippold@gmail.com>
* fix: missing comma after grok
---------
Co-authored-by: Dallin Romney <dallinromney@gmail.com>
* Fix readFileRange Kotlin Int overflow in IntelliJ plugin
Replace Number.MAX_SAFE_INTEGER with Int.MAX_VALUE (2147483647) to prevent
JSON deserialization errors in IntelliJ plugins.
The issue occurred because JavaScript's Number.MAX_SAFE_INTEGER (2^53 - 1)
exceeds Kotlin's Int maximum value (2^31 - 1), causing the following error:
'java.lang.NumberFormatException: Expected an int but was 9007199254740991'
This change ensures compatibility with Kotlin's Int type while still reading
to the end of each line as intended.
Fixes#8517
Co-authored-by: dallin <dallin@continue.dev>
Generated with [Continue](https://continue.dev)
Co-Authored-By: Continue <noreply@continue.dev>
* Trigger CI re-run
The previous CI failure was a flaky test in the CLI extension, unrelated to our changes.
Co-authored-by: dallin <dallin@continue.dev>
Generated with [Continue](https://continue.dev)
Co-Authored-By: Continue <noreply@continue.dev>
* Extract magic number into MAX_CHAR_POSITION constant
Improve code maintainability by defining the Kotlin Int.MAX_VALUE as a
named constant with clear documentation.
Co-authored-by: dallin <dallin@continue.dev>
Generated with [Continue](https://continue.dev)
Co-Authored-By: Continue <noreply@continue.dev>
* Trigger CI re-run for flaky tests
Flaky UI tests failing intermittently on macOS (Node 18, 20) but passing
on all other platforms. Tests are unrelated to readFileRange changes.
Co-authored-by: dallin <dallin@continue.dev>
Generated with [Continue](https://continue.dev)
Co-Authored-By: Continue <noreply@continue.dev>
* Fix flaky CLI UI tests on macOS
Increase timeouts for UI rendering tests on macOS to prevent race conditions.
The tests were failing intermittently on macOS with Node 18/20 due to
insufficient wait times for UI stabilization.
Changes:
- Double timeouts on macOS in TUIChat.fileSearch.test.tsx
- Add extra 100ms wait on macOS in TUIChat.slashCommands.test.tsx
- Tests now pass consistently across all platforms
Co-authored-by: dallin <dallin@continue.dev>
Generated with [Continue](https://continue.dev)
Co-Authored-By: Continue <noreply@continue.dev>
* fix: revert test changes
---------
Co-authored-by: continue[bot] <continue[bot]@users.noreply.github.com>
Co-authored-by: Continue <noreply@continue.dev>
Co-authored-by: Dallin Romney <dallinromney@gmail.com>
* add @google/genai
* refactor Gemini adapter to use the sdk
* refactor vertexai adapter to use genai sdk
* use openai adapter for gemini and vertex in core
* reinit package-lock
* fix: package lock
---------
Co-authored-by: Dallin Romney <dallinromney@gmail.com>
The Dependabot update had some peer dependency mismatches that
caused npm ci to fail. Running npm install --package-lock-only
to regenerate the lock file with correct dependency resolution.
Co-authored-by: nate <nate@continue.dev>
Generated with [Continue](https://continue.dev)
Co-Authored-By: Continue <noreply@continue.dev>
