sequence-llm 0.2.1

A sequence-based LLM orchestration framework

Sequence-LLM

Sequence-LLM is a terminal-first orchestration tool for running local large language models through llama.cpp (llama-server) with automatic server lifecycle management, profile-based switching, and reproducible workflows.

It removes the need to manually start servers, remember commands, or write shell scripts when working with multiple models.

Sequence-LLM works with any hardware supported by llama.cpp — CPU, CUDA GPUs, ROCm, Metal, and more.

Cross-platform: Windows, Linux, macOS.

---

Why Sequence-LLM

Running local models often involves:

• Manually starting and stopping servers
• Remembering model paths and ports
• Managing multiple configurations
• Writing ad-hoc scripts to switch models
• Repeating setup across machines

Sequence-LLM solves this by providing:

• Named model profiles
• Automatic start and shutdown of servers
• Interactive chat interface
• Consistent configuration across machines
• Deterministic, script-free workflows

---

Who Is This For

• Developers running multiple local models
• AI engineers building local pipelines
• Researchers comparing architectures
• Self-hosting enthusiasts
• GPU workstation users
• CLI-first workflows

If you use tools like llama.cpp, with upcoming support for Ollama, LM Studio, or custom scripts - Sequence-LLM simplifies the workflow.

---

Features

• Interactive CLI built with Typer and Rich
• Profile-based model switching (/brain, /coder, etc.)
• Automatic shutdown of previous server before starting a new one
• Health checking with readiness polling
• Context-window safety guard (prevents overflow / crashes)
• Cross-platform process management using subprocess and psutil
• OS-aware configuration directory creation
• Conversation history management
• Status panel showing active model and server info
• First-run configuration wizard

---

Hardware Support

Sequence-LLM does not perform inference itself.

It orchestrates llama-server, meaning it works with:

• CPU inference
• NVIDIA CUDA GPUs
• AMD ROCm GPUs
• Apple Metal
• Any backend supported by llama.cpp

---

Comparison with Other Tools

Tool	Primary Focus	Sequence-LLM Advantage
Ollama	Easy installs	Multi-model orchestration workflow
LM Studio	GUI	Lightweight CLI automation
Raw llama.cpp	Flexible	No manual scripts needed
Open-WebUI	Web UI	Minimal overhead terminal workflow

Sequence-LLM sits between simplicity and flexibility.

---

Installation

Requirements

• Python 3.9+
• llama-server binary from llama.cpp

Install from PyPI:

pip install sequence-llm

---

Quick Start

Run the CLI:

seq-llm

On first launch, a configuration file is created automatically.

Config locations:

• Windows: %APPDATA%\sequence-llm\config.yaml
• Linux: ~/.config/sequence-llm/config.yaml
• macOS: ~/Library/Application Support/sequence-llm/config.yaml

---

Configuration Example

llama_server: "/path/to/llama-server"

defaults:
  threads: 6
  threads_batch: 8
  batch_size: 512

profiles:
  brain:
    name: "Brain Model"
    model_path: "/path/to/model.gguf"
    system_prompt: "/path/to/system.txt"
    port: 8081
    ctx_size: 16384
    temperature: 0.7

  coder:
    name: "Coder Model"
    model_path: "/path/to/coder.gguf"
    system_prompt: "/path/to/coder.txt"
    port: 8082
    ctx_size: 32768
    temperature: 0.3

---

CLI Usage

/status   → show active model and server status
/brain    → switch to brain profile
/coder    → switch to coder profile
/clear    → clear conversation history
/quit     → stop server and exit

Typing any text sends a message to the active model.

---

Example Workflow

1. Start CLI
2. Automatically load default model
3. Switch between models using commands
4. Chat interactively without restarting processes manually

---

Architecture

User → CLI → ServerManager → llama-server → Model
           ↑
        Config + Profiles

Core components:

• CLI - interactive interface and command routing
• Server Manager - lifecycle control of llama-server
• API Client - communication with local inference server
• Config System - YAML-based profiles and defaults

---

Roadmap

Planned evolution:

• v0.3 — Multi-model named workflows
• v0.4 — TUI interface
• v0.5 — Hardware auto-optimization
• v1.0 — Production stability

---

For Development and Contributors

Clone repository:

git clone https://github.com/Ananay28425/Sequence-LLM.git
cd Sequence-LLM
pip install -e .

Run tests:

pytest -v

---

License

AGPL-3.0 License. See LICENSE file for details.

---

Contributing

Pull requests and issues are welcome.

GitHub: https://github.com/Ananay28425/Sequence-LLM

---

Sequence-LLM provides a lightweight and predictable way to manage local LLM workflows from the terminal.

Sequence-LLM - Orchestrate LLM workflows with ease.