|
|
"""Quick test of diverse generation.""" |
|
|
import json |
|
|
import random |
|
|
import os |
|
|
from dotenv import load_dotenv |
|
|
load_dotenv() |
|
|
|
|
|
import cohere |
|
|
|
|
|
client = cohere.ClientV2(api_key=os.getenv("COHERE_API_KEY")) |
|
|
|
|
|
|
|
|
category = "company.tools_config" |
|
|
industry = "Series A fintech building a neobank" |
|
|
persona = "a growth lead obsessed with metrics" |
|
|
situation = "debugging why a campaign tanked" |
|
|
tone = "frustrated" |
|
|
|
|
|
prompt = f"""You are a world-class creative writer generating training data for an AI memory routing system. |
|
|
|
|
|
Create a completely unique, realistic conversation between {persona} at a {industry} and their AI marketing assistant. |
|
|
|
|
|
Context: They are {situation}. The tone is {tone}. |
|
|
|
|
|
CATEGORY TO DEMONSTRATE: {category} |
|
|
The conversation should involve tool setup, integrations, APIs, or workflow automation. |
|
|
|
|
|
CREATIVE FREEDOM: |
|
|
- Invent specific, realistic details (names, numbers, dates, products) |
|
|
- The conversation can start anywhere - mid-thought, mid-project, mid-crisis |
|
|
- Vary structure dramatically |
|
|
- Include natural speech patterns |
|
|
- Make it feel like eavesdropping on a real conversation |
|
|
|
|
|
The ONLY hard requirement: the conversation must clearly demonstrate {category}. |
|
|
|
|
|
Output as JSON: |
|
|
{{"scenario_id": "unique_id", "conversation": [{{"role": "user", "content": "..."}}, {{"role": "assistant", "content": "..."}}], "labels": {{"categories": ["{category}"]}}, "metadata": {{"primary_category": "{category}", "industry": "{industry}"}}}}""" |
|
|
|
|
|
print("Sending request...") |
|
|
response = client.chat( |
|
|
messages=[{"role": "user", "content": prompt}], |
|
|
temperature=0.95, |
|
|
model="command-r-plus-08-2024", |
|
|
response_format={"type": "json_object"} |
|
|
) |
|
|
|
|
|
content = response.message.content[0].text |
|
|
print("\n=== RAW RESPONSE ===") |
|
|
print(content[:500]) |
|
|
|
|
|
data = json.loads(content) |
|
|
print("\n=== PARSED ===") |
|
|
print(f"Categories: {data.get('labels', {}).get('categories', [])}") |
|
|
conv = data.get("conversation", []) |
|
|
if conv: |
|
|
for i, turn in enumerate(conv[:4]): |
|
|
if isinstance(turn, dict): |
|
|
print(f"\n[{turn.get('role', 'unknown')}]: {turn.get('content', '')[:150]}...") |
|
|
else: |
|
|
print(f"\n[turn {i}]: {str(turn)[:150]}...") |
|
|
|
|
|
|
