🚀 4CBON2 — 12-Agent Cognitive Ecosystem (Gemini Frontier Research Edition)

Powered by the Google Generative AI API with source-grounded AI, mathematics and science research databases.

🔑 Bring your own key. Paste a Google AI Studio API key once in any tab (Get a free key at https://aistudio.google.com/apikey). It is kept in this browser session's memory only — never stored on the server or written to disk, and isolated from other visitors. Default model: gemini-3.6-flash. Runtime data is ephemeral and resets on Space restart.

Upload .txt, .pdf, or .docx files to the knowledge base.

Upload file

Status

Frontier Research Question

Ask an ambitious AI, mathematics, or science question. The system automatically routes it to the strong local domain databases and can retrieve current records from pertinent scholarly databases.

Your Question

Source-grounded Answer

Retrieval Report

Multi-Agent Orchestration with 12 specialists powered by Google Gemini.

Execution Log & Output

AI Rewriter — 16-Layer Pipeline

Paste an AI-generated answer and optionally describe its goal. The runtime executes each layer in order, scores the answer before and after, preserves the three L9 Socratic questions for the next run, and includes three free runs per session.

AI-generated answer

Context / goal (optional)

Ready. No pipeline run yet.

Layer outputs — L0 → P → W → LX → LA → LC → L1 → L2 → L3 → LP → L4 → LR → L6 → L7 → L8 → L9 → L10

L0 — Interpretation Engine

P — Parsing Layer

W — World Model Layer

LX — Reality Adjudication

LA — Adversarial Countermodel

LC — Compression Integrity

L1 — Hypothesis Engine

L2 — Evaluation Layer

L3 — Rewrite Planner

LP — Policy Translation

L4 — Finalization Engine

LR — Regret Layer

L6 — Trace Memory

L7 — Curriculum Generator

L8 — Identity Model

L9 — Socratic Integrity Engine

L10 — Synthesis/Audit

Final L10 Audit

Copy All report

Request a Custom Agent

Tell the 4CBON2 team what you need. Requests may be free or paid depending on complexity. Submitting prepares an email to mohamtur1@gmail.com; review the draft and press Send in your email client.

Your email (required)

Agent description (required)

Additional notes (optional)

Task Memory Visualization

Visualize your agent task history with interactive Plotly charts.

Metrics:

Task response length over time
Subtask distribution
Common words in task goals

Status

View all agent conversation histories.

Click refresh to load.

About this Space

4CBON2 is a 12-agent cognitive ecosystem with source-grounded research retrieval and a 16-layer answer-rewriting pipeline.

Component	Value
LLM	Google Generative AI API (`google-genai`)
Default model	`gemini-3.6-flash`
Pipeline order	`L0 → P → W → LX → LA → LC → L1 → L2 → L3 → LP → L4 → LR → L6 → L7 → L8 → L9 → L10`
Vector store	ChromaDB — `all-MiniLM-L6-v2 (ONNX, semantic)`
Runtime data	`./data` (ephemeral)
Source notebook	`4CBOn2_Gemini2c.ipynb`

🔑 Your API key

Create one free at Google AI Studio.
It is held in this browser session's memory only — never written to disk, never logged.
Each request is isolated via contextvars, so concurrent visitors cannot use your key or your quota.
Optionally set GEMINI_API_KEY as a Space secret to run without pasting a key. Be aware that on a public Space this makes your key and quota available to every visitor.

⚠️ Cost awareness

One AI Rewriter run makes roughly 20 Gemini calls (17 layers plus a 3-call median score before and after). One Agent Mode run can make 15–50 calls depending on the plan. gemini-3.6-flash is billed per token on your key.

🗂 Ephemeral storage

Uploaded documents, cached live records, agent history and task memory live under ./data and reset whenever the Space restarts or sleeps. The curated AI / Mathematics / Science research indexes are reseeded automatically on every boot.

Built with Gradio logo