Σ-Avatarus for TiDB — FDL Agent Kernel with Vector Memory

Inspi## What is it

Σ-Avatarus / FDL Agent Kernel is an open agent platform for building multi-scenario AI agents with an ethical loop. In this version, we use TiDB Cloud (Serverless + Vector Search) as a state store, "memory," and agent step log.

Why

• Transparent decision logic (FDL: thesis ↔ antithesis ↔ synthesis).
• Ethical output verification (LexiconGuard / CounterFrame).
• TiDB integration: vector memory, full-text, state audit log.
• Network fairness: AGPL-3.0 guarantees the return of improvements.

What it does

• Builds semantic loops and request processing routes (multi-step agents).
• Maintains a step-by-step state log in TiDB: messages, tools, observations.
• Stores "vector memory" (embeddings) in TiDB Vector for contextual retrieval.
• Supports security pipelines: moderation, anti-manipulation, counter-frames.

Architecture

• Core: Python (plugin-based agents) + FDL protocols.
• Data: TiDB Cloud (tables agent_runs, agent_steps, mem_vectors).
• Embeddings: any compatible provider; vectors — in TiDB Vector.
• CI: GitHub Actions (lint/SPDX/tests).

Challenges

• Balancing ethics and performance.
• Replicability of FDL scenarios across domains.
• Reliability and traceability of agent steps (via TiDB).

Achievements

• FDL core MVP + routing plugins.
• Unified Ethics and Legal Package (AGPL + DCO + Governance).
• MOU templates for fast pilots.

Repositories

• GPT-N7Δ+: https://github.com/NgoiSigma/gpt-n7delta-plus
• FDL Agent Kernel: https://github.com/NgoiSigma/fdl-agent-kernel
• FDL CODEX: https://github.com/NgoiSigma/--FDL-CODEX

License

AGPL-3.0-or-later.ration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Σ-Avatarus for TiDB — FDL Agent Kernel with Vector Memory

Built With

• ai/ml;
• github
• python;-tidb-cloud-(serverless
• search);
• vector