hexai: switch ollama models (minimax-m2.1, qwen3-coder:480b)

Set the main ollama model to minimax-m2.1:cloud and the in-editor completion
model to qwen3-coder:480b-cloud (temperature 0.0 for deterministic completion),
and drop the commented-out alternative-model lists.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

1 files changed, 4 insertions, 16 deletions

diff --git a/hexai/config.toml b/hexai/config.toml
index 6382b2c..db86860 100644
--- a/hexai/config.toml
+++ b/hexai/config.toml
@@ -16,24 +16,12 @@ slashcommand_dir = "~/.cursor/commands"
 name = "ollama"
 
 [ollama]
-model       = "gemma4:31b-cloud"
-# Smaller Gemma variants on Ollama Cloud — uncomment one if gemma4:31b-cloud
-# becomes too slow:
-# model     = "gemma3:27b"
-# model     = "gemma3:12b"
-# model     = "gemma3:4b"
+model       = "minimax-m2.1:cloud"
 base_url    = "https://ollama.com"
 temperature = 0.2
 
-# In-code auto-completion uses qwen3-coder-next on Ollama Cloud — a
-# coding-specialized model offering a good balance of speed and quality.
+# Use a low-latency model for in-editor completion.
 [[models.completion]]
 provider    = "ollama"
-model       = "qwen3-coder-next:cloud"
-# Other cloud variants — uncomment one if qwen3-coder-next:cloud is not
-# the right fit:
-# model     = "ministral-3:14b-cloud"
-# model     = "devstral-small-2:24b-cloud"
-# model     = "gemma4:31b-cloud"
-# model     = "gemma3:12b"
-temperature = 0.2
+model       = "qwen3-coder:480b-cloud"
+temperature = 0.0