ResearchOps v1.1 (Expert Edition)

Serverless Multi-Modal Intelligence Pipeline for R&D & Process Engineering

ResearchOps is an automated intelligence system designed for R&D Managers and Engineers. It acts as a "Second Brain", ingesting technical content from various sources (Web, PDF, Images) and synthesizing it into actionable insights using advanced LLMs.

🚀 Features

• Omni-Channel Ingest (WF1): Send URLs, Text, Images (Screenshots), or PDFs via Telegram. The system automatically analyzes, categorizes, and summarizes them into Markdown.
• Strategic Monitor (WF2): Monthly hybrid intelligence report combining your field captures (Internal) with fresh web/academic search results (External).
• Deep Research (WF3): Autonomous multi-agent research system ("Tri-Force") that combines Perplexity (Market/News), Gemini+Tavily (Tech/Engineering), and Claude Opus (Strategic Synthesis) to produce master-level reports on complex topics.
• Zero-Infra: Runs entirely on GitHub Actions. No servers to manage.
• Expert Analysis: Prompts tuned for Industrial ROI, Feasibility, and Innovation.

🛠 Architecture

graph TB
    subgraph "Input"
        TG[Telegram] -->|Make.com| GH[GitHub Issue]
    end

    subgraph "Workflows"
        GH -->|Label: veille| WF1[WF1: Ingest]
        GH -->|Label: research| WF3[WF3: Deep Research]
        CRON[Schedule] -->|Monthly| WF2[WF2: Monitor]
    end

    subgraph "Intelligence"
        WF1 --> GEM[Gemini Flash]
        WF2 --> CLAUDE[Claude Sonnet]
        WF3 --> TRI[Tri-Force Agents]
    end

    GEM -->|Markdown| REPO[(Content Repo)]
    CLAUDE -->|Report| REPO
    TRI -->|Master Report| REPO

Loading

📂 Project Structure

• src/: Python source code for workflows.
  • wf1_ingest.py: Daily ingestion logic.
  • wf2_monitor.py: Monthly strategic monitoring.
  • wf3_triforce.py: Deep research orchestration.
  • agents/: AI Agents for WF3 (Perplexity, Gemini, Claude).
  • utils/: Helper modules.
• config/: Configuration files.
• content/: Stored intelligence (Markdown files).
• reports/: Generated monthly reports.
• research/: Deep research master reports.
• docs/: Detailed documentation.

🚦 Setup & Usage

1. Initial Setup

1. Secrets: Configure API Keys (Google, Anthropic, Perplexity, Tavily, Telegram) in GitHub Secrets.
2. Make.com: Set up the Telegram to GitHub bridge (see docs/make_setup.md).

2. How to Run the Workflows

Workflow	Trigger Type	How to Launch
WF1: Ingest	Automatic	Send a message/file to your Telegram Bot. Make.com creates an issue with label `veille`.
WF2: Monitor	Scheduled / Manual	Runs automatically on the 1st of the month. Can be triggered manually via GitHub Actions tab (`Run workflow`).
WF3: Research	Manual (Issue)	Create a GitHub Issue with label `research`. • Title: Research Topic (e.g., "Solid State Batteries") • Body: (Optional) Specific questions, context, or constraints.

3. Detailed Triggers

• WF1 (Daily Watch): Triggered by any Issue labeled `veille`.
• WF2 (Monthly Report): Triggered by CRON `0 6 1 * *` or `workflow_dispatch`.
• WF3 (Deep Research): Triggered by any Issue labeled `research`.
  • Pro Tip: Use the Issue Body to guide the agents (e.g., "Focus on European market", "Ignore patents before 2020").

📚 Documentation

• Make.com Setup Guide
• Workflow 1: Ingestion Details
• Workflow 2: Monitoring Details
• Workflow 3: Deep Research Details

👤 Author

Younes AJEDDIG Ph.D - R&D Manager / Process Modeling & Simulation Expert / Data Engineer & scientist / Scientific developer