KnowCode/aimodels.yaml at main · deepakdgupta1/KnowCode · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
natural_language_models:
  - name: glm-5
    provider: z-ai
    # The proxy's key, not the provider's. This traffic goes to the LiteLLM
    # proxy, which holds the provider keys itself.
    api_key_env: LITELLM_MASTER_KEY
    rpm_free_tier_limit: 99999
    rpd_free_tier_limit: 99999

embedding_models:
  - name: voyage-code-3
    provider: voyageai
    api_key_env: LITELLM_MASTER_KEY
    tokens_free_tier_limit: 200000000

# Reranking calls VoyageAI directly rather than through the proxy (BL-44), so
# this entry names the provider's own key, and doctor checks that it is set.
reranking_models:
  - name: rerank-2.5
    provider: voyageai
    api_key_env: VOYAGE_API_KEY_1

eval_models:
  - name: voyage-code-3
    provider: voyageai
    #   api_key_env: VOYAGE_API_KEY_1
    tokens_free_tier_limit: 200000000

# Configuration
config:
  sufficiency_threshold: 0.8 # Configurable threshold for local-first answering
  # Generated only after the machine-verification gate passes. Empty is fail-closed.
  local_answer_task_types: []
  routing_quality_floor: 0.9
  hybrid_alpha: 0.2
  reranker_top_k_multiplier: 5