codegen/.env.example at main · Supernova-45/codegen · 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
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
# =====================================
# Core Runtime (required)
# =====================================
# OpenAI-compatible endpoint used for code generation.
CODEGEN_API_KEY=replace_me
CODEGEN_MODEL=meta-llama/llama-3.1-8b-instruct
CODEGEN_BASE_URL=https://api.openai.com/v1

# =====================================
# Runtime Tuning (optional)
# =====================================
# Minimum spacing between requests (seconds); useful for reducing 429s.
# CODEGEN_MIN_REQUEST_INTERVAL_S=0
# Retry budget for transient 429/5xx failures.
# CODEGEN_MAX_REQUEST_ATTEMPTS=4
# Max output tokens for code generation and test generation calls.
# CODEGEN_MAX_TOKENS=10000
# CODEGEN_TESTGEN_MAX_TOKENS=3000

# =====================================
# Split Test-Generation Model (optional)
# =====================================
# If unset, test generation uses the same model as code generation.
# CODEGEN_TESTGEN_API_KEY=replace_me
# CODEGEN_TESTGEN_BASE_URL=https://api.groq.com/openai/v1
# CODEGEN_TESTGEN_MODEL=llama-3.3-70b-versatile
# CODEGEN_TESTGEN_TEMPERATURE=0.2
# CODEGEN_TESTGEN_REQUEST_TIMEOUT_S=90

# =====================================
# API Key Pools (optional, for sharding/rotation)
# =====================================
# Option 1: point to env-var names that each contain one key.
# CODEGEN_API_KEY_ENV_POOL=GROK_KEY_1,GROK_KEY_2,GROK_KEY_3,GROK_KEY_4,GROK_KEY_5,GROK_KEY_6
# GROK_KEY_1=replace_me
# GROK_KEY_2=replace_me
# GROK_KEY_3=replace_me
# GROK_KEY_4=replace_me
# GROK_KEY_5=replace_me
# GROK_KEY_6=replace_me
#
# Option 2: provide keys directly in one variable.
# CODEGEN_API_KEY_POOL=key1,key2,key3,key4,key5,key6

# =====================================
# Modal Endpoint (optional)
# =====================================
# Used by scripts/deploy_modal_qwen.sh and model_profiles.yaml.
# MODAL_OPENAI_BASE_URL=https://<your-endpoint>.modal.run
# MODAL_BASE_URL=https://<your-endpoint>.modal.run/v1
# MODAL_API_KEY=dummy
# MODAL_QWEN_CHEAP_MODEL=Qwen/Qwen2.5-7B-Instruct
# MODAL_QWEN_GPU=L4
# MODAL_QWEN_MAX_CONTAINERS=6
# MODAL_QWEN_BUFFER_CONTAINERS=1

# =====================================
# Cross-Model Profile Variables (optional)
# =====================================
# Used by scripts/run_model_matrix.py with configs/model_profiles.yaml.
# GROQ_BASE_URL=https://api.groq.com/openai/v1
# GROQ_API_KEY=replace_me
# GROQ_LLAMA_TEST_MODEL=llama-3.3-70b-versatile
# OPENAI_API_KEY=replace_me
# OPENAI_MID_MODEL=gpt-4.1-mini
# OPENAI_LARGE_MODEL=gpt-5
# GOOGLE_OPENAI_BASE_URL=https://<your-google-openai-compat-endpoint>/openapi
# GOOGLE_OPENAI_API_KEY=replace_me
# GOOGLE_LARGE_MODEL=gemini-3-pro-preview

# =====================================
# Local Gemini Vertex Proxy (optional)
# =====================================
# Used by scripts/run_google_humaneval_oneshot.sh and servers/gemini_vertex_openai_server.py.
# GCP_PROJECT_NAME=your-gcp-project
# GEMINI_VERTEX_MODEL=gemini-3-pro-preview
# GEMINI_VERTEX_LOCATION=global
# GEMINI_VERTEX_HOST=127.0.0.1
# GEMINI_VERTEX_PORT=8010
# Optional thinking override for Gemini proxy (unset = model default).
# GEMINI_VERTEX_THINKING_BUDGET=0
# Optional override for one-shot Google wrapper token cap.
# GOOGLE_CODEGEN_MAX_TOKENS=50000

# Legacy compatibility note:
# Older CLARIFYCODE_* names are still accepted by runtime code but are deprecated.