IdeaHub Integration Guide

March 3, 2026 · View on GitHub

NeuriCo can automatically fetch research ideas from IdeaHub and convert them into minimal YAML format (v1.1).

Overview

IdeaHub is a platform for sharing and discovering research ideas. This integration allows you to:

Fetch ideas from IdeaHub URLs
Convert them to minimal NeuriCo format (preserves only provided info)
Submit them directly - agent will research missing details
Run autonomous experiments with agent-driven resource discovery

Setup

Prerequisites

OpenAI API Key (for GPT-4 conversion)
- Get your API key from: https://platform.openai.com/api-keys
- Add to your .env.local or .env file:
```
OPENAI_API_KEY=sk-your-key-here
```

Dependencies (already included in pyproject.toml)

uv sync  # Installs requests, beautifulsoup4, openai

Usage

Basic Usage

Fetch an idea from IdeaHub:

python src/cli/fetch_from_ideahub.py https://hypogenic.ai/ideahub/idea/HGVv4Z0ALWVHZ9YsstWT

What happens (v1.1):

📥 Fetches the idea content from IdeaHub (title, description, references)
🤖 Uses GPT-4o-mini to format as minimal YAML (no hallucination)
💾 Saves only provided information - no invented datasets/methods
✅ Agent will research missing details when you run the experiment

Example output:

================================================================================
IdeaHub to NeuriCo Converter
================================================================================
📥 Fetching idea from IdeaHub...
   URL: https://hypogenic.ai/ideahub/idea/HGVv4Z0ALWVHZ9YsstWT

✓ Found idea: Do LLMs differentiate epistemic belief from non-epistemic belief?

🤖 Converting to NeuriCo format using GPT-4...
   Calling GPT-4 API...
   ✓ Conversion complete

✅ Idea saved to: ideas/do_llms_differentiate_epistemic_belief.yaml

To submit this idea:
  python src/cli/submit.py ideas/do_llms_differentiate_epistemic_belief.yaml

================================================================================
Done!
================================================================================

Auto-Submit

Fetch and automatically submit the idea:

python src/cli/fetch_from_ideahub.py https://hypogenic.ai/ideahub/idea/HGVv4Z0ALWVHZ9YsstWT --submit

This will:

Fetch and convert the idea
Validate it against the schema
Submit it to NeuriCo
Create a GitHub repository (if GitHub integration enabled)
Return the idea ID for running

Example output:

...
📤 Submitting idea to NeuriCo...

✓ Idea submitted successfully: do_llms_differentiate_epistemic_belief_20250103_120000_abc123de

To run this research:
  python src/core/runner.py do_llms_differentiate_epistemic_belief_20250103_120000_abc123de

Custom Output Path

Save to a specific location:

python src/cli/fetch_from_ideahub.py \
  https://hypogenic.ai/ideahub/idea/HGVv4Z0ALWVHZ9YsstWT \
  --output my_ideas/llm_beliefs.yaml

How It Works

1. Content Extraction

The script fetches the IdeaHub page and extracts:

Title: Main research question
Description: Detailed explanation and context
Tags: Research domains and topics
Author: Original idea creator
References: Cited papers and resources

2. GPT-4 Conversion

GPT-4 is prompted to convert the IdeaHub content into structured YAML following the NeuriCo schema:

idea:
  title: "Extracted or refined title"
  domain: artificial_intelligence  # Inferred from content
  hypothesis: "Clear, testable hypothesis"

  background:
    description: "Context and motivation"
    papers: [...]  # Extracted references

  methodology:
    approach: "Proposed experimental approach"
    steps: [...]
    baselines: [...]
    metrics: [...]

  constraints:
    compute: cpu_only  # For AI research
    budget: "\$50-150"  # Typical API costs

  expected_outputs: [...]
  evaluation_criteria: [...]

GPT-4's Role:

Domain Classification: Infers appropriate domain from tags and content
Hypothesis Extraction: Formulates testable hypothesis from description
Methodology Design: Proposes experimental steps, baselines, and metrics
Constraint Estimation: Sets realistic compute, time, and budget constraints
Output Specification: Defines expected results and evaluation criteria

3. Validation & Saving

The converted YAML is:

Validated against the schema
Enhanced with metadata (source, source_url)
Saved with a sanitized filename derived from the title

Examples

Example 1: AI/LLM Research

IdeaHub URL: https://hypogenic.ai/ideahub/idea/HGVv4Z0ALWVHZ9YsstWT

IdeaHub Content:

Title: "Do LLMs differentiate epistemic belief from non-epistemic belief?"
Description: Research on whether LLMs exhibit distinct types of beliefs
Tags: Psychology, LLM behavior

Converted YAML:

idea:
  title: "Evaluating Epistemic vs Non-Epistemic Belief Differentiation in LLMs"
  domain: artificial_intelligence

  hypothesis: |
    LLMs demonstrate measurable differences in representing epistemic beliefs
    (knowledge-based) versus non-epistemic beliefs (religious, moral),
    similar to human cognitive patterns.

  methodology:
    approach: "Comparative prompt-based evaluation"
    steps:
      - "Design prompts testing epistemic beliefs (factual knowledge)"
      - "Design prompts testing non-epistemic beliefs (values, preferences)"
      - "Run across multiple LLMs (GPT-4, Claude, Gemini)"
      - "Analyze response patterns and confidence levels"
      - "Compare with human baseline from Vesga et al. (2025)"

    baselines:
      - "Human belief differentiation patterns from psychology research"
      - "Zero-shot vs few-shot prompting"

    metrics:
      - "Belief type classification accuracy"
      - "Confidence level differences"
      - "Response consistency across similar prompts"

Example 2: Complete Workflow

# 1. Fetch idea from IdeaHub
python src/cli/fetch_from_ideahub.py \
  https://hypogenic.ai/ideahub/idea/ABC123 \
  --submit

# Output: idea_id_20250103_120000_abc123de

# 2. (Optional) Add resources to workspace
cd workspace/idea-id-20250103-120000-abc123de
# Add datasets, papers, etc.
git add . && git commit -m "Add resources" && git push
cd ../..

# 3. Run the research
python src/core/runner.py idea_id_20250103_120000_abc123de

Troubleshooting

"OPENAI_API_KEY not found"

Solution: Add your API key to .env.local or .env:

# Create .env.local if it doesn't exist
cp .env.example .env.local

# Edit .env.local
nano .env.local

# Add:
OPENAI_API_KEY=sk-your-key-here

"Error fetching URL"

Possible causes:

Network connectivity issues
Invalid IdeaHub URL
Page structure changed

Solution:

Verify the URL is correct and accessible
Check internet connection
Try again after a few seconds

"Generated YAML may have issues"

Cause: GPT-4 occasionally generates invalid YAML syntax

Solution:

Check the saved YAML file for syntax errors
Manually fix any issues (usually missing quotes or colons)

Re-run the submission manually:

python src/cli/submit.py ideas/generated_file.yaml

Rate Limiting

Issue: GPT-4 API rate limits

Solution:

Wait a few seconds between requests
Use GPT-4 Turbo (higher rate limits)
Check your OpenAI account quotas

Advanced Usage

Custom Conversion Prompts

If you need to customize how ideas are converted, you can modify the GPT-4 prompt in src/cli/fetch_from_ideahub.py:

# Line ~140 in fetch_from_ideahub.py
prompt = f"""You are an expert research assistant...
# Customize this prompt to change conversion behavior

Batch Processing

Fetch multiple ideas:

#!/bin/bash
# fetch_multiple.sh

urls=(
  "https://hypogenic.ai/ideahub/idea/ABC123"
  "https://hypogenic.ai/ideahub/idea/DEF456"
  "https://hypogenic.ai/ideahub/idea/GHI789"
)

for url in "${urls[@]}"; do
  python src/cli/fetch_from_ideahub.py "$url"
  sleep 2  # Avoid rate limits
done

Manual Review

Always review converted ideas before running:

# Fetch without auto-submit
python src/cli/fetch_from_ideahub.py <url>

# Review the generated YAML
cat ideas/generated_file.yaml

# Edit if needed
nano ideas/generated_file.yaml

# Submit manually
python src/cli/submit.py ideas/generated_file.yaml

Best Practices

1. Review Before Running

GPT-4 does a good job converting ideas, but always review:

✅ Check the hypothesis is testable
✅ Verify methodology is realistic
✅ Ensure constraints match your resources
✅ Confirm metrics are appropriate

2. Add Context

After conversion, you can enhance the idea:

Add specific datasets you have access to
Include references to relevant papers
Adjust time/budget constraints to your needs
Specify additional evaluation criteria

3. Use Resources

If the IdeaHub idea references specific papers or datasets:

Fetch the idea with --submit
Navigate to the workspace
Add the referenced papers/datasets
Commit and push before running

cd workspace/<repo-name>
mkdir docs && cp ~/papers/*.pdf docs/
git add . && git commit -m "Add referenced papers" && git push
cd ../..
python src/core/runner.py <idea_id>

API Costs

Typical costs per conversion:

GPT-4 Turbo: ~$0.02-0.10 per idea
GPT-4: ~$0.05-0.20 per idea

Cost optimization:

Use GPT-4 Turbo (cheaper, faster)
Batch process during off-peak hours
Cache conversions (script saves YAML, reuse if possible)

Future Enhancements

Potential improvements to the IdeaHub integration:

Support for fetching multiple ideas at once
Integration with IdeaHub user accounts
Automatic tagging and categorization
Direct submission to IdeaHub from NeuriCo results
Community sharing of converted ideas

FAQ

Q: Does this work with private IdeaHub ideas?

A: Currently only public ideas are supported. Private ideas require authentication, which is not yet implemented.

Q: Can I convert ideas from other platforms?

A: The script is designed for IdeaHub, but you could adapt it for other platforms by modifying the content extraction logic.

Q: What if the conversion doesn't match what I want?

A: Simply edit the generated YAML file before submitting. The GPT-4 conversion is a starting point that you can refine.

Q: Are the original IdeaHub authors credited?

A: Yes! The conversion includes the author name and source URL in the metadata. Always respect original authorship and licensing.

Q: Can I use a different model instead of GPT-4?

A: Yes, modify the script to use Claude or Gemini. GPT-4 is used by default for its strong instruction following.

Last Updated: 2025-11-03 Maintained by: ChicagoHAI