fix: do not override Ollama model num_ctx configuration

[roomote]

Description

This PR fixes an issue where the native Ollama handler was overriding user-configured num_ctx values in their Ollama model configurations.

Problem

In version 3.25.16, when the Ollama integration was switched to use the native API instead of the OpenAI compatibility layer, the implementation started explicitly setting num_ctx to modelInfo.contextWindow. This overrode any custom num_ctx value that users had configured in their Ollama models.

For users who had specifically configured their models with smaller context sizes to fit within their available memory (e.g., setting num_ctx to 16384), this caused the models to use the full context window (e.g., 1048576), leading to out-of-memory errors.

Solution

This fix removes the explicit num_ctx setting from the options passed to Ollama, allowing the service to use the model's configured value as intended. This restores the behavior from version 3.25.15 and earlier.

Changes

• Removed num_ctx: modelInfo.contextWindow from the options in both createMessage and completePrompt methods
• Added tests to verify that num_ctx is not being set in the options passed to Ollama

Testing

• All existing tests pass
• Added new tests specifically to verify num_ctx is not overridden
• Manually verified that Ollama now uses the model's configured context size

Fixes #7159

---

Important

Fixes issue where num_ctx was overridden in Ollama models, restoring user-configured values and preventing out-of-memory errors.

• Behavior:
  • Removes num_ctx: modelInfo.contextWindow from createMessage and completePrompt in native-ollama.ts.
  • Restores behavior from version 3.25.15, allowing user-configured num_ctx values.
• Testing:
  • Adds tests in native-ollama.spec.ts to verify num_ctx is not set in options for createMessage and completePrompt.
  • All existing tests pass, ensuring no regression.
• Misc:
  • Fixes issue Ollama model num_ctx is being ignored in the newest version #7159.

^{This description was created by for df38bfe. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Reviewing my own code because apparently I trust no one, not even myself.

[roomote]

Good fix removing the num_ctx override. However, could we consider adding documentation about how the Ollama context configuration works with this handler? It might help future developers understand why we're not setting this value explicitly.

[roomote]

These comments are identical in both methods. Is this intentional? We could consider extracting this behavior note to a class-level comment or constant to avoid duplication and make it clear this is a deliberate pattern across all Ollama API calls.

[roomote]

Nice test coverage for verifying num_ctx isn't set! Could we also add an integration test that verifies the handler works correctly when Ollama models have different configured context sizes? This would help catch any future regressions.

[daniel-lxs]

This fix is incomplete. It removes the num_ctx override (good) but leaves a display issue: Roo Code shows the model's default context window, not the user's configured num_ctx value.

The proper fix needs to parse the actual num_ctx from Ollama's parameters field to display the correct context size. See issue #7159 for details.

[hannesrudolph]

Related to: #7797