mdkrushr/vera-ai

By mdkrushr

•Updated 7 months ago

Persistent Memory Proxy for Ollama - Gives your AI conversations lasting memory

Image

2.1K

Overview Tags

mdkrushr/vera-ai repository overview

⁠Vera-AI - Persistent Memory Proxy for Ollama

Vera (Latin): True — True AI Memory

⁠What is Vera-AI?

Vera-AI is a transparent proxy for Ollama that adds persistent memory using Qdrant vector storage. It sits between your AI client and Ollama, automatically augmenting conversations with relevant context from previous sessions.

Every conversation is remembered.

⁠How It Works

Client → Vera-AI:11434 → Ollama:11434 → Response
              ↓
         Qdrant:6333
              ↓
         Memory Storage

Request comes from client to Vera-AI
Query semantic memory from Qdrant
Forward augmented request to Ollama
Store conversation turn in Qdrant
Return response to client

⁠Quick Start

# Pull from Docker Hub
docker pull mdkrushr/vera-ai:latest

# Run with Ollama and Qdrant
docker run -d \
  --name VeraAI \
  --restart unless-stopped \
  --network host \
  -e APP_UID=$(id -u) \
  -e APP_GID=$(id -g) \
  -e TZ=America/Chicago \
  -v ./config/config.toml:/app/config/config.toml:ro \
  -v ./prompts:/app/prompts:rw \
  -v ./logs:/app/logs:rw \
  mdkrushr/vera-ai:latest

⁠Prerequisites

Service	Description
Ollama	LLM inference server (`curl https://ollama.ai/install.sh \| sh`)
Qdrant	Vector database (`docker run -d -p 6333:6333 qdrant/qdrant`)
Docker	Container runtime

⁠Install Ollama

curl https://ollama.ai/install.sh | sh
ollama pull snowflake-arctic-embed2  # Embedding model
ollama pull llama3.1                   # Chat model

⁠Features

Feature	Description
🧠 Persistent Memory	Conversations stored in Qdrant
📅 Daily Curation	Automatic memory cleanup
🔍 4-Layer Context	System + semantic + recent + current
👤 Configurable UID/GID	Match container user to host
🌍 Timezone Support	Scheduler runs in your timezone

⁠Configuration

⁠Environment Variables

Variable	Default	Description
`APP_UID`	`999`	Container user ID
`APP_GID`	`999`	Container group ID
`TZ`	`UTC`	Container timezone
`VERA_DEBUG`	`false`	Enable debug logging

⁠config.toml

[general]
ollama_host = "http://127.0.0.1:11434"
qdrant_host = "http://127.0.0.1:6333"
qdrant_collection = "memories"
embedding_model = "snowflake-arctic-embed2"
debug = false

[layers]
semantic_token_budget = 25000
context_token_budget = 22000
semantic_search_turns = 2
semantic_score_threshold = 0.6

[curator]
run_time = "02:00"
full_run_time = "03:00"
full_run_day = 1
curator_model = "llama3.1"

⁠API Endpoints

Endpoint	Method	Description
`/`	GET	Health check
`/api/chat`	POST	Chat completion (with memory)
`/api/tags`	GET	List models
`/curator/run`	POST	Trigger curator manually

⁠Verify Installation

# Health check
curl http://localhost:11434/
# Expected: {"status":"ok","ollama":"reachable"}

# Test chat
curl -X POST http://localhost:11434/api/chat \
  -H "Content-Type: application/json" \
  -d '{"model":"llama3.1","messages":[{"role":"user","content":"hello"}],"stream":false}'

⁠Source Code

GitHub: https://github.com/speedyfoxai/vera-ai⁠

⁠License

MIT License - Brought to you by SpeedyFoxAi

Tag summary

Recent tags

Content type

Image

Digest

sha256:ae30b62a1…

Size

90.6 MB

Last updated

7 months ago

docker pull mdkrushr/vera-ai