[Laurianti]

Settings passed through ChatOptions.RawRepresentationFactory were partly overwritten by AsChatClient:

• MaxTokens was reset to 256 when MaxOutputTokens wasn't set
• Temperature, TopP and TopK were reset to the defaults
• Seed was replaced with a random one
• A sampling pipeline other than DefaultSamplingPipeline was replaced

These now fall back to the raw settings, the same way the other sampling settings already did. Values set on ChatOptions still take priority.

Added LLamaExecutorChatClientTests, which don't need a model.

[Copilot]

Copilot review overview

🟡 Changes recommended

Preserve raw MaxTokens, including the documented unlimited value -1.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1

Open (1)

• Preserve raw MaxTokens=-1 instead of overriding with 256 · New

What changed in this PR

This PR updates AsChatClient to preserve settings from RawRepresentationFactory while honoring explicit ChatOptions values.

Changes:

• Preserves raw token and sampling settings.
• Retains custom sampling pipelines.
• Adds model-independent regression tests.

A moderate issue remains: raw MaxTokens = -1 is still overwritten with 256 when MaxOutputTokens is unset.

File	Description
LLama/​Extensions/​LLamaExecutorExtensions.cs	Updates inference parameter merging.
LLama.Unittest/​LLamaExecutorChatClientTests.cs	Adds coverage for raw settings and option precedence.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[martindevans]

Thanks for fixing this. Looks good to merge once the tests are completed :D