LatentGate

Process Locally. Send Smart. Pay Less.

A VL-JEPA-inspired pipeline that compresses images, text, conversations, and RAG documents locally via Ollama, then sends only compact payloads to any LLM API — every saving measured with a real tokenizer and checked for lost facts.

Use in Claude | Quick Start | Python API | REST API | Monitoring | Load Testing | Deployment | AI Tools | Benchmarks | Contributing

Traditional:  Image -> Cloud LLM (1,200 tokens) -> Answer
LatentGate:   Image -> Local Ollama (FREE) -> Cloud LLM (200 tokens) -> Answer

uvx --from "latent-gate[mcp,tokens]" latent-gate --optimizer-benchmark

/plugin marketplace add KathanModh259/latent-gate
/plugin install latent-gate@latent-gate

claude mcp add latent-gate -- uvx --from "latent-gate[mcp,tokens]" latent-gate-mcp

{
  "mcpServers": {
    "latent-gate": {
      "command": "uvx",
      "args": ["--from", "latent-gate[mcp,tokens]", "latent-gate-mcp"]
    }
  }
}

# Core install
pip install latent-gate

# With MCP server (for Claude Desktop, Cursor, Cline, etc.)
pip install latent-gate[mcp]

# With API server (for web applications)
pip install latent-gate[api]

# With video processing
pip install latent-gate[video]

# With embedding-based similarity (more accurate selective decoding)
pip install latent-gate[embeddings]

# With LangChain integration
pip install latent-gate[langchain]

# With AWS Bedrock support
pip install latent-gate[bedrock]

# Exact token counts via tiktoken (otherwise a calibrated ~6%-error estimate)
pip install latent-gate[tokens]

# With all features
pip install latent-gate[all]

ollama pull llava:7b      # Vision model (required for image queries)
ollama pull llama3:8b     # Text model (required for text compression & prediction)

chmod +x scripts/quickstart.sh
./scripts/quickstart.sh

# Image query
latent-gate photo.jpg "What is in this image?" --provider ollama -v

# Text compression
latent-gate --text "Your long prompt here..." --provider ollama -v

# Text from file
latent-gate --text-file prompt.txt --provider openai -v

# Image + Text combined
latent-gate photo.jpg "Analyze" --text "Extra context..." -v

# Full JSON output
latent-gate photo.jpg "Describe" --json -v

# Compress only, deterministically (no LLM call, ~1ms, reproducible)
cat prompt.txt | latent-gate --text-file - --compress-only --deterministic --level balanced

# Reproduce the token-savings benchmark (no Ollama needed)
latent-gate --optimizer-benchmark

# Production benchmark
latent-gate --benchmark --benchmark-output reports/benchmark.json

# Start API server (requires: pip install latent-gate[api])
latent-gate-api

set LATENTGATE_ALLOWED_IMAGE_ROOTS=C:\safe-images;D:\datasets
latent-gate-api

latent-gate --benchmark --json

from latent_gate import LatentGatePipeline, PipelineConfig

config = PipelineConfig(
    vision_model="llava:7b",
    predictor_model="llama3:8b",
    remote_provider="openai",
    remote_model="gpt-4o-mini",
)

with LatentGatePipeline(config) as pipeline:
    result = pipeline.query("photo.jpg", "What is in this image?")

    print(result["answer"])
    print(f"Tokens sent: ~{result['tokens_estimated']}")
    print(f"Timing: {result['timing']}")

# Long prompt compression
result = pipeline.query_text("Your 500-word prompt here...", mode="auto")

# Conversation history compression
messages = [
    {"role": "user", "content": "Help me with Kubernetes setup"},
    {"role": "assistant", "content": "Sure! What's your target configuration?"},
    {"role": "user", "content": "3 nodes, t3.large, us-east-1 with autoscaling"},
]
result = pipeline.query_conversation(messages, "Now give me the setup commands")

# RAG document compression
documents = ["doc1 text...", "doc2 text...", "doc3 text..."]
result = pipeline.query_documents(documents, "How do I implement JWT refresh?")

# Universal (auto-detect input type)
result = pipeline.query_universal(text="Explain this code...", image="screenshot.png")

# Sequential with selective decoding (skips redundant API calls)
results = pipeline.query_batch(image_paths, "Describe each scene")

# Parallel processing
results = pipeline.query_batch(image_paths, "Describe each scene", parallel=True, max_workers=4)

# Text batch
results = pipeline.query_batch_texts(text_list, question="Summarize each")

# Stream image query
for token in pipeline.query_stream("photo.jpg", "Describe this"):
    print(token, end="", flush=True)