Hugging Face's logo

Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
Back to feed
mallocode200 
posted an update Jun 21
Post
5239
Hello everyone, happy to share with you my experimentation of a Deep Research Assistant, using 7 agents and a quality assurance pipeline :

🤖 What makes this special:

✅ Agent-Based Architecture - 7 specialised AI agents working together:
- Planner Agent - Strategic search planning
- Search Agent - Multi-source web research
- Writer Agent - Comprehensive report generation
- Evaluator Agent - Automatic quality assessment
- Optimiser Agent - Iterative improvement when needed
- Email Agent - Professional report delivery
- Clarifier Agent - Interactive query refinement

✅ Quality Assurance Pipeline - Every report is scored (1-10) and automatically improved if it scores below 7/10

✅ Multiple Research Modes - From quick queries to deep, clarification-driven analysis

✅ Production-Ready - Deployed on Hugging Face Spaces with comprehensive documentation

🔧 Technical Stack:
- Frontend: Gradio with theme-adaptive UI
- Backend: OpenAI Agents framework
- Integration: SendGrid for email delivery
- Deployment: Containerised with full CI/CD pipeline
- Tracing: Full OpenAI trace integration for transparency

💡 Real-World Impact:
This isn't just another AI tool - it's a complete research workflow that delivers publication-quality reports with built-in fact-checking and optimisation. Perfect for consultants, researchers, analysts, and anyone who needs reliable, comprehensive research.

🚀 Key Features:
- Automatic quality evaluation and improvement
- Email delivery of formatted reports
- Interactive clarification for targeted results
- Full traceability and audit trails
- Professional documentation and deployment guides
- Built with modern AI engineering principles: modular design, quality assurance, and production deployment in mind.
- The entire codebase is organised with clean separation of concerns - each agent has a specific role, making it maintainable and extensible.

mallocode200/Deep_Research_Assistant
VIATEUR-AI

Thank you very much for your insightful feedback and great questions! I’m glad you see value and innovation in the project.

How agents communicate: Our agents share information via prompt chaining combined with API calls, using a shared memory system so each agent can access the outputs of the previous ones and continue the workflow seamlessly.

Evaluation metrics: The Evaluator Agent uses multiple criteria including coherence, completeness, factual accuracy, and language quality. It also cross-checks facts against trusted sources where possible.

Typical runtime: For standard queries, a full research cycle takes about 3-5 minutes. More complex or deep-dive queries involving clarifications can take 10-15 minutes depending on the interaction.

Performance on diverse topics: We have tested on a wide range of subjects including highly specialized domains, and the system performs well especially when interactive clarifications help narrow down the focus early on.

Open source plans: Yes, we plan to open source the project within the next few months and are eager to receive community feedback and contributions.

I hope this gives a clearer picture of the project. I’d be happy to continue discussing ideas like adding visual reasoning components or adaptive prompt modes to improve efficiency and depth.

Thanks again for your interest and valuable questions!

sulaman786
•
edited Jun 22

This is truly next-level agent orchestration — the modular QA pipeline and the use of clearly defined, role-specific agents are incredibly well thought out. 👏

I recently came across CodeLibrary.ai, a curated directory of reusable LLM agent patterns — including planner chains, fallback strategies, memory handling techniques, and evaluation flows. What you’ve built here aligns perfectly with the kind of structured, production-ready agent workflows that CodeLibrary highlights.

It would be fantastic to see this framework featured as a reference use case — it’s one of the most compelling examples of how MCP-style architecture can be applied in real-world scenarios. 🚀

·
mallocode200

Thanks so much for the kind words!

I'd be delighted to collaborate and have the Deep Research Assistant featured as a use case. It's exciting to see the alignment between your work on LLM agent patterns and the architecture I've developed.

Let me know what you have in mind for the next steps.
Looking forward to it!

greattkiffy

You want to allow users to select a model of their choice and BYOAPI _key to avoid the 429 errors from openAI

·
mallocode200

Great point, thank you! You're right, letting users bring their key would solve the 429 errors and make it much more scalable. I'll look into adding that.

nickg68

Thanks for sharing the details on this. Very cool! You've inspired me to look more into deep research projects.

In this post