mirror of
https://github.com/ggozad/oterm.git
synced 2026-11-03 11:51:24 +01:00
The context window only reached the summarize capability when a local server reported one, so cloud models always fell back to keeping the last 20 messages. get_agent now resolves the model before building the agent and reads the window from its profile. The summary run built its own agent with no model settings, so the chat's temperature, thinking switch and openai-compat chat template switch did not apply. The summary now uses the chat's parameters with thinking off and no max_tokens, so a user cap cannot cut it short. |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| test_agent.py | ||
| test_capabilities.py | ||
| test_cli.py | ||
| test_config.py | ||
| test_context_window.py | ||
| test_provider_settings.py | ||
| test_providers.py | ||
| test_providers_ollama.py | ||
| test_store.py | ||
| test_store_upgrades.py | ||
| test_types.py | ||
| test_utils.py | ||