Skip to content

Commit 6ddd83c

Browse files
committed
feat(ccv3): optimize for cheapest Fireworks model + secure env
- Use minimax-m2p1 as default model (cheapest on Fireworks) - Add .env.example with placeholder values (no secrets) - Add .gitignore to prevent committing .env files - Never commit API keys - follow secure SDLC
1 parent 34d82ee commit 6ddd83c

3 files changed

Lines changed: 38 additions & 8 deletions

File tree

opc/ccv3/.env.example

Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,14 @@
1+
# CCv3 Hackathon Environment Variables
2+
# Copy to .env and fill in your values: cp .env.example .env
3+
4+
# MongoDB Atlas - REQUIRED FOR FINALISTS
5+
MONGODB_URI="mongodb+srv://USER:PASSWORD@cluster.mongodb.net/?appName=ccv3"
6+
7+
# Jina AI - v3 embeddings with task adapters
8+
JINA_API_KEY="your_jina_api_key"
9+
10+
# Fireworks AI - using minimax-m2p1 (cheapest)
11+
FIREWORKS_API_KEY="your_fireworks_api_key"
12+
13+
# Galileo AI - RAG Triad evaluation
14+
GALILEO_API_KEY="your_galileo_api_key"

opc/ccv3/.gitignore

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,15 @@
1+
# Environment files with secrets
2+
.env
3+
.env.local
4+
.env.hackathon
5+
*.env
6+
7+
# Python
8+
__pycache__/
9+
*.pyc
10+
*.pyo
11+
.pytest_cache/
12+
13+
# IDE
14+
.vscode/
15+
.idea/

opc/ccv3/inference.py

Lines changed: 9 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -32,19 +32,20 @@
3232

3333

3434
# Task → Model mapping (PRD Section 3.2)
35+
# OPTIMIZED FOR COST - using minimax-m2p1 as default cheap model
3536
TaskType = Literal["planning", "analysis", "coding", "patching", "cheap", "strong"]
3637

3738
TASK_MODEL_MAP: dict[TaskType, str] = {
38-
"planning": "accounts/nvidia/models/nemotron-3-8b-chat-v1", # NVIDIA Nemotron
39-
"analysis": "accounts/fireworks/models/qwen2-72b-instruct", # Strong analysis
40-
"coding": "accounts/fireworks/models/qwen2-72b-instruct", # Strong coding
41-
"patching": "accounts/fireworks/models/llama-v3p1-70b-instruct", # Strong patching
42-
"cheap": "accounts/nvidia/models/nemotron-3-8b-chat-v1", # Cost optimized
43-
"strong": "accounts/fireworks/models/qwen2-72b-instruct", # Quality optimized
39+
"planning": "accounts/fireworks/models/minimax-m2p1", # CHEAPEST - use for planning
40+
"analysis": "accounts/fireworks/models/minimax-m2p1", # CHEAPEST - use for analysis
41+
"coding": "accounts/fireworks/models/minimax-m2p1", # CHEAPEST - for hackathon demo
42+
"patching": "accounts/fireworks/models/minimax-m2p1", # CHEAPEST - for hackathon demo
43+
"cheap": "accounts/fireworks/models/minimax-m2p1", # CHEAPEST - explicit cheap
44+
"strong": "accounts/fireworks/models/llama-v3p1-70b-instruct", # Only use when really needed
4445
}
4546

46-
# Default model
47-
DEFAULT_MODEL = "accounts/fireworks/models/llama-v3p1-70b-instruct"
47+
# Default model - CHEAPEST for hackathon
48+
DEFAULT_MODEL = "accounts/fireworks/models/minimax-m2p1"
4849

4950

5051
class InferenceLLM:

0 commit comments

Comments
 (0)