Smriti AI: Memory Layer for Frozen Small Models
Collection
Training-free external memory, semantic recall, graph recall, deletion, and production gates for frozen Gemma, Qwen, and small models. • 6 items • Updated
Text Generation
English
smriti-memory-ai
smriti-ai
gemma
gemma-4
qwen
qwen2.5
gemma-memory
memory
agent-memory
long-term-memory
external-memory
training-free
frozen-model
inference-time-augmentation
retrieval-augmented-generation
rag
semantic-search
knowledge-graph
identity-continuity
small-language-model
small-language-models
ai-agent
Smriti AI for Gemma 4 Memory
Smriti AI for Gemma 4 Memory is a lightweight discovery card for the Smriti AI inference-time memory layer. It helps users looking for Gemma memory, Gemma 4 long-term memory, agent memory, semantic recall, graph recall, and training-free memory augmentation find the main Smriti AI project.
Important
This repository is not an official Gemma checkpoint, and it is not fine-tuned. Use google/gemma-4-E2B-it when you need the base checkpoint itself. The base model remains frozen. Smriti AI stores memory outside the weights and injects retrieved context at inference time.
Use
Use the main model-style repository and package:
pip install "smriti-memory-ai[ml]==1.0.9"
from smriti import SmritiAILite, MemPalaceLite
For Hugging Face Inference Endpoints, use luciferai-devil/smriti-ai with:
BASE_MODEL_ID=google/gemma-4-E2B-it
Related assets
- Main model wrapper:
luciferai-devil/smriti-ai - Demo Space:
luciferai-devil/smriti-ai-demo - Benchmark dataset:
luciferai-devil/smriti-ai-benchmarks - GitHub:
https://github.com/Luciferai04/smriti-ai - PyPI:
smriti-memory-ai
Claim boundary
Current official benchmark evidence includes Gemma 4 plus Qwen 2.5, Phi-3 mini, and Llama 3.2 1B real-model shards with provenance. Smriti AI improves memory access; it does not change model weights.