https://github.com/perseus-computing-llc/perseus-rapid-agent
Perseus Memory Agent: Gemini-powered agent with persistent project context using Elastic + Engram-rs. Google Cloud Rapid Agent Hackathon 2026 submission.
https://github.com/perseus-computing-llc/perseus-rapid-agent
Last synced: 2 days ago
JSON representation
Perseus Memory Agent: Gemini-powered agent with persistent project context using Elastic + Engram-rs. Google Cloud Rapid Agent Hackathon 2026 submission.
- Host: GitHub
- URL: https://github.com/perseus-computing-llc/perseus-rapid-agent
- Owner: Perseus-Computing-LLC
- License: mit
- Created: 2026-06-06T22:04:01.000Z (22 days ago)
- Default Branch: main
- Last Pushed: 2026-06-13T02:21:25.000Z (16 days ago)
- Last Synced: 2026-06-17T06:26:18.698Z (12 days ago)
- Language: Python
- Homepage:
- Size: 8.88 MB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Agents: AGENTS.md
Awesome Lists containing this project
README
# Perseus Rapid Agent — Elastic Partner Track
**One env var. Cloud to self-hosted. Same agent, zero code changes.**
A Gemini-powered agent with persistent project memory. Switches between Elastic Cloud (ELSER semantic search, ES|QL, Kibana) and Engram-rs (local SQLite+FTS5) by changing one environment variable. Built for the [Google Cloud Rapid Agent Hackathon](https://rapid-agent.devpost.com/) — Elastic Partner Track.
[](./LICENSE)
[]()
## The Dual-Backend Architecture
```
MemoryBackend (abstract)
/ \
Elastic Cloud (MCP) Engram-rs (local)
· ELSER semantic search · SQLite + FTS5
· ES|QL reflect queries · Sub-millisecond recall
· Kibana dashboard · Zero cloud deps
· Managed, scalable · Self-hosted, MIT
```
| | Elastic (Cloud) | Engram-rs (Local) |
|---|---|---|
| **Search** | ELSER semantic | FTS5 keyword |
| **Queries** | ES|QL structured | SQL |
| **Dashboard** | Kibana | None |
| **Deps** | Elastic Cloud account | Zero |
| **Swap cost** | `MEMORY_BACKEND=elastic` | `MEMORY_BACKEND=engram` |
## Why Semantic Search Matters
```
Query: "how to build fast"
Memory: "we use Rust for speed"
Engram (Keyword): ✗ MISS — "build fast" ≠ "Rust for speed"
Elastic (ELSER): ✓ MATCH (0.84) — ELSER understands Rust is fast
```
Hybrid search finds what keywords miss. The dual-backend architecture lets you choose.
## Quick Start
```bash
# Clone and install
git clone https://github.com/tcconnally/perseus-rapid-agent.git
cd perseus-rapid-agent
pip install -r requirements.txt
# Elastic Cloud (partner track)
export MEMORY_BACKEND=elastic
export ELASTIC_CLOUD_ID=...
python -m agent.main
# Engram-rs (self-hosted, zero cloud)
export MEMORY_BACKEND=engram
python -m agent.main
```
## Demo: 3-Minute Walkthrough
1. **Session 1** — Developer tells agent about their project stack
2. **Session 2** — Agent recalls project context without being reminded
3. **Session 3** — Agent compounds knowledge, notices a pattern from past sessions
4. **Backend Switch** — Same code, swap Elastic → Engram-rs, memory persists
## Hackathon Details
- **Platform:** Google Cloud Rapid Agent Hackathon on Devpost
- **Track:** Elastic Partner Track
- **License:** MIT
---
[Website](https://perseus.observer/rapid-agent/) · [Engram-rs](https://github.com/tcconnally/engram-rs)