gurix/Meister-Eder
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

introduce separate SIMPLE_MODEL config for lightweight tasks

Browse Source 
Add a distinct simple_model field alongside ai_model so operators can
route cheap, simple tasks (e.g. email-label translation) to a low-cost
model while keeping the strong model for parent conversations. The two
models can be from different providers (e.g. Gemini + Haiku). If
SIMPLE_MODEL is unset, falls back to AI_MODEL with a warning.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
gurix
2026-02-23 22:18:28 +01:00
co-authored by Claude Sonnet 4.6
parent 14fde88c89
commit 602b8eddc3
4 changed files with 36 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.env.example
+12 -5
View File
@@ -6,14 +6,21 @@
# ---------------------------------------------------------------
# ---------------------------------------------------------------
# AI Model (via litellm — supports any provider)
# AI Models (via litellm — supports any provider)
# ---------------------------------------------------------------
# Use litellm model strings: "<provider>/<model-name>"
# Examples:
# anthropic/claude-opus-4-6 (default)
# openai/gpt-4o
# gemini/gemini-2.0-flash
#
# AI_MODEL — primary model for conversation with parents (complex reasoning).
# SIMPLE_MODEL — lightweight model for simple tasks, e.g. email-label translation.
# Can be from a different provider than AI_MODEL.
# If unset, falls back to AI_MODEL with a warning in the logs.
#
# Examples — mixing providers:
# AI_MODEL=gemini/gemini-3-pro-preview + SIMPLE_MODEL=gemini/gemini-3-flash
# AI_MODEL=anthropic/claude-opus-4-6 + SIMPLE_MODEL=anthropic/claude-haiku-4-5-20251001
# AI_MODEL=gemini/gemini-3-pro-preview + SIMPLE_MODEL=anthropic/claude-haiku-4-5-20251001
AI_MODEL=anthropic/claude-opus-4-6
SIMPLE_MODEL=anthropic/claude-haiku-4-5-20251001
# Extended thinking — Anthropic models only (leave unset to disable).
# Enables a reasoning phase before the model's reply, which improves