antoinezambelli/forgePublic

A reliability layer for self-hosted LLM tool-calling. You give forge a set of tools; the model calls whichever it wants in whatever order. Workflow structure is opt-in — required_steps, prerequisites, and terminal_tool let you constrain the loop when you need to, but forge's guardrails (rescue parsing, retry nudges, response validation) apply with zero required steps too.

Forge takes an 8B local model from single digits to 84% across forge's 26-scenario v0.7.0 eval suite — and even lifts Sonnet 4.6 from 85% to 98% on the same workload (Anthropic numbers measured in v0.6.0; not re-run in v0.7.0 since the cost is non-trivial).

What forge isn't:

• Not an agent orchestrator. Forge sits inside one agentic loop and makes its tool calls reliable. Multi-agent graphs, DAG planners, and cross-agent coordination are out of scope.
• Not a coding harness. Forge is domain-agnostic. If you're building a coding agent (or already using one like opencode, aider, Cline), proxy mode lifts your existing harness with forge's guardrails — no rewrite.

Three ways to use it:

• Proxy server — Drop-in proxy (forge-proxy from the standalone distribution, or python -m forge.proxy from the Python package) speaking both the OpenAI chat-completions and Anthropic Messages (/v1/messages) APIs, sitting between any client and a local model server. Point OpenAI-compatible tools (opencode, Continue, aider) or Claude Code at it and forge applies guardrails transparently — the client thinks it's talking to a smarter model. Most popular entry point.

• WorkflowRunner — Define tools, pick a backend, run structured agent loops. Forge manages the full lifecycle: system prompts, tool execution, context compaction, and guardrails. SlotWorker adds priority-queued access to a shared inference slot with auto-preemption — for multi-agent architectures where specialist workflows share a GPU slot. Best when you're building on forge directly.

• Guardrails middleware — Use forge's reliability stack (composable middleware) inside your own orchestration loop. You control the loop; forge validates responses, rescues malformed tool calls, and enforces required steps.

Supports generic OpenAI-compatible endpoints, Ollama, llama-server (llama.cpp), Llamafile, vLLM, and Anthropic as backends.

Forge Proxy is a self-contained developer sidecar: point an OpenAI- or Anthropic-compatible client at it to add Forge guardrails without rewriting the client or integrating the Python library. The command bundles Forge, its private Python runtime, and the Anthropic SDK, so the host does not need Python or pip. It does not install a backend executable, model, GPU stack, service, credentials, or client configuration.

Install the latest verified standalone Proxy release:

Linux and macOS:

curl -fsSL https://raw.githubusercontent.com/antoinezambelli/forge/main/install.sh | sh

Windows PowerShell:

irm https://raw.githubusercontent.com/antoinezambelli/forge/main/install.ps1 | iex

Open a refreshed terminal, then create and validate a profile:

forge-proxy init
forge-proxy check

See Forge Proxy Installation for supported platforms, exact-version installation, profiles, updates, recovery, and uninstall.

Use the Python package for WorkflowRunner, guardrails middleware, development, or a Python-managed Proxy. It requires:

• Python 3.12+
• A running LLM backend (see below)

pip install forge-guardrails                # core only
pip install "forge-guardrails[anthropic]"   # + Anthropic client

The Python package intentionally does not install a global forge-proxy command. Run its Proxy implementation with python -m forge.proxy; the standalone installer above is the sole owner of the forge-proxy command and its update/uninstall lifecycle.

For development:

git clone https://github.com/antoinezambelli/forge.git
cd forge
pip install -e ".[dev]"

llama-server (recommended — top 10 eval configs all run on llama-server):

# Install from https://github.com/ggml-org/llama.cpp/releases
llama-server -m path/to/Ministral-3-8B-Instruct-2512-Q8_0.gguf --jinja -ngl 999 --port 8080

Ollama (alternative — easier setup, slightly weaker on harder workloads):

# Install from https://ollama.com/download
ollama pull ministral-3:8b-instruct-2512-q4_K_M

Anthropic (API, no local GPU needed):

pip install -e ".[anthropic]"
export ANTHROPIC_API_KEY=sk-...

See Backend Setup for full instructions and Model Guide for which model fits your hardware.

Start llama-server however you normally do (e.g. in a separate shell):

llama-server -m path/to/Ministral-3-8B-Instruct-2512-Q8_0.gguf --jinja -ngl 999 --port 8080

Then the Python you'll run (e.g. from another shell):

import asyncio
from pydantic import BaseModel, Field
from forge import (
    Workflow, ToolDef, ToolSpec,
    WorkflowRunner, LlamafileClient,
    ContextManager, TieredCompact,
)

def get_weather(city: str) -> str:
    return f"72°F and sunny in {city}"

class GetWeatherParams(BaseModel):
    city: str = Field(description="City name")

workflow = Workflow(
    name="weather",
    description="Look up weather for a city.",
    tools={
        "get_weather": ToolDef(
            spec=ToolSpec(
                name="get_weather",
                description="Get current weather",
                parameters=GetWeatherParams,
            ),
            callable=get_weather,
        ),
    },
    required_steps=[],
    terminal_tool="get_weather",
    system_prompt_template="You are a helpful assistant. Use the available tools to answer the user.",
)

async def main():
    client = LlamafileClient(
        gguf_path="path/to/Ministral-3-8B-Instruct-2512-Q8_0.gguf",
        mode="native",
        recommended_sampling=True,
    )
    ctx = ContextManager(strategy=TieredCompact(keep_recent=2), budget_tokens=8192)
    runner = WorkflowRunner(client=client, context_manager=ctx)
    await runner.run(workflow, "What's the weather in Paris?")

asyncio.run(main())

For multi-step workflows, multi-turn conversations, and backend auto-management, see the User Guide. If you're building a long-running session (CLI, chat server, voice assistant), see the long-running session advisory for important guidance on filtering transient messages.

Upgrading from pre-0.9: 0.9.0 is one intentional breaking Proxy update. See Migrating to Forge 0.9 for the complete old/new/action table.

Drop-in proxy that sits between any client and a local model server, speaking both the OpenAI chat-completions API and the Anthropic Messages API (/v1/messages). Point your client at the proxy (e.g. http://localhost:8081/v1) and forge applies its guardrails transparently — the client thinks it's talking to a smarter model.

This is the path for using forge with an existing harness (opencode, Continue, aider, Cline, anything that speaks the OpenAI chat-completions schema — or Claude Code, which speaks the Anthropic Messages API). No Python rewrite. Reasoning replay defaults to none: Forge still captures reasoning for observability, but keeps it out of backend-facing history on later turns — the most token-efficient policy, and statistically indistinguishable from replay-all on the eval suite (see reasoning-replay results). Use --reasoning-replay keep-last to replay only the latest reasoning block, or --reasoning-replay full for the historical replay-all behavior.

# External llama-server — you manage the backend, forge proxies it
python -m forge.proxy --backend-url http://localhost:8080 --backend llamaserver --port 8081

# External Anthropic-shaped downstream
python -m forge.proxy --backend-url https://gateway.example --backend anthropic --model claude-route --port 8081

# Managed mode — forge spawns or attaches to the backend, then starts the proxy
python -m forge.proxy --backend llamaserver --gguf path/to/model.gguf --port 8081

# Managed vLLM — pass a model directory or HF repo id via --model-path
python -m forge.proxy --backend vllm --model-path /path/to/awq-dir --port 8081

Then configure your client to use http://localhost:8081/v1 as the API base URL.

Claude Code: the proxy also serves the Anthropic Messages API on POST /v1/messages, so you can point Claude Code at a forge-guarded local model — set ANTHROPIC_BASE_URL=http://localhost:8081 and ANTHROPIC_AUTH_TOKEN=anything for the claude process. See Using forge with Claude Code for the full setup (native-vs-prompt FC, Anthropic-shape downstreams, cache_control).

Backend compatibility:

• Managed mode spawns llama-server, llamafile, or vLLM, or attaches to an existing Ollama daemon. Supported selectors are llamaserver, llamafile, ollama, and vllm (use --gguf for the GGUF-based backends, --model-path for vLLM, or --model for Ollama). Stopping Forge unloads the selected Ollama model but does not stop or own the daemon.
• External mode uses --backend-url. Omission or --backend openai selects generic OpenAI compatibility; the complete explicit selector set is openai, anthropic, llamaserver, llamafile, ollama, and vllm. The specialized selectors choose the corresponding adapter and metadata behavior. On the generic OpenAI/llama profile, --model is a fallback used only when the inbound request omits model; an inbound value wins. On Anthropic and vLLM profiles, --model pins the wire identity. For unpinned vLLM, the first inference request discovers the served identity. --budget-tokens independently supplies only a reporting denominator and never suppresses required unpinned identity discovery.

On tool-bearing inference requests, forge applies (in order):

1. Response validation — each tool call in the model's response is checked against the tools array in the request. Calls to unknown tool names or with malformed shapes are caught before the response returns to your client.
2. Rescue parsing — when the model emits tool calls in the wrong format (JSON in a code fence, Mistral's [TOOL_CALLS]name{args}, Qwen's <tool_call>...</tool_call> XML), forge extracts the structured call and re-emits it in the canonical OpenAI tool_calls schema. Biggest practical lift for Mistral-family models.
3. Retry loop with error tracking — if validation fails, forge retries inference up to --max-retries (default 3) with a corrective tool-result message on the canonical channel, rather than returning a malformed response. From your client's perspective the proxy looks like a single request that just took a few extra ms.
4. Optional synthetic respond tool injection — --inject-respond-tool opts into a synthetic respond tool when tools are present. It defaults off. When enabled, the call is stripped from the outbound response and becomes normal text. See ADR-013 for the rationale.

Tool-free requests bypass validation, rescue, and retry and go directly to the selected backend adapter.

Proxy mode is single-shot per request; some forge features need multi-turn workflow state that the OpenAI chat-completions schema doesn't carry:

• Prerequisite enforcement and step-ordering — these need a workflow definition spanning turns. Available in WorkflowRunner.
• Context compaction and session memory — proxy mode never compacts or deletes caller history. The preserved llama/OpenAI adapter normalization may merge consecutive visible same-role messages for backend template compatibility; that is distinct from budget-driven compaction. Managing the rolling window is the client's job.
• Unmanaged backend operation — metadata and --budget-tokens are reporting-only. The operator owns model allocation/swaps, overflow rejection, readiness, and backend failures. Managed Proxy modes retain backend, manual, forge-full, and forge-fast allocation behavior.

Read-only backend metadata is forwarded only on GET /health, /v1/health, /v1/models, /models, and /props. Use /forge/health for Forge liveness. Forwarding is transparent: if the selected backend does not implement a route, Forge returns its status unchanged, including Ollama's 404 for /health. /forge/usage reports one last-completed process-local snapshot or 204; it is not a live meter, ledger, or persistent session API. Forge Proxy is a per-operator sidecar and does not authenticate callers; --backend-api-key authenticates Forge to the backend, not callers to Forge.

For the full guardrail surface, use WorkflowRunner directly. The proxy trades depth for "use forge with your existing setup, no rewrite."

Managed Ollama also rejects process/KV controls it cannot apply: cache_type_k, cache_type_v, n_slots, and kv_unified.

You can run the forge proxy as a Docker container.

Build the image:

docker build -t forge-proxy .

Run the container:

# Connect to an external backend (e.g. vLLM hosted on the same machine)
docker run -p 8081:8081 forge-proxy --backend-url http://host.docker.internal:8000 --backend vllm --budget-tokens 8192

Note: If your backend is running on localhost of the host machine, use http://host.docker.internal:PORT (on macOS/Windows) or the host's IP address to allow the container to reach it.

See Backend Setup for installation and Model Guide for which model to pick.

python -m pytest tests/ -v --tb=short

python -m pytest tests/ --cov=forge --cov-report=term-missing

For proxy changes, also run the deterministic proxy smoke test and the manual real-backend sanity check described in Contributing.

Scenarios measuring how reliably a model + backend combo navigates multi-step tool-calling workflows, with a baseline tier and an advanced_reasoning tier for top-end separation. See Eval Guide for the current scenario inventory and full CLI reference.

The released run-level outcome corpus is available as the Forge eval dataset on Hugging Face.

# llama-server (start in another terminal first; see Eval Guide)
python -m tests.eval.eval_runner --backend llamafile --llamafile-mode prompt --gguf "path/to/Ministral-3-8B-Instruct-2512-Q8_0.gguf" --runs 10 --stream --verbose

# Batch eval (JSONL output, automatic resume)
python -m tests.eval.batch_eval --config all --runs 50

# Reports — ASCII table by default; --html / --markdown export views
python -m tests.eval.report eval_results.jsonl
python -m tests.eval.report eval_results.jsonl --html docs/results/dashboard.html
python -m tests.eval.report eval_results.jsonl --markdown docs/results/

src/forge/
  __init__.py          # Public API exports
  errors.py            # ForgeError hierarchy
  server.py            # setup_backend(), ServerManager, BudgetMode
  core/
    messages.py        # Message, MessageRole, MessageType, MessageMeta
    workflow.py        # ToolSpec, ToolDef, ToolCall, TextResponse, Workflow
    inference.py       # run_inference() — shared front half (compact, fold, validate, retry)
    runner.py          # WorkflowRunner — the agentic loop
    slot_worker.py     # SlotWorker — priority-queued slot access
    steps.py           # StepTracker
  guardrails/
    guardrails.py      # Guardrails facade — applies the full stack in foreign loops
    nudge.py           # Nudge dataclass
    response_validator.py  # ResponseValidator, ValidationResult
    step_enforcer.py   # StepEnforcer, StepCheck
    error_tracker.py   # ErrorTracker
  clients/
    base.py            # ChunkType, StreamChunk, LLMClient protocol
    ollama.py          # OllamaClient (native FC)
    llamafile.py       # LlamafileClient (native FC or prompt-injected)
    openai_compat.py   # OpenAICompatClient (generic OpenAI-shaped endpoints)
    vllm.py            # VLLMClient (vLLM-specific identity and response handling)
    anthropic.py       # AnthropicClient (frontier baseline)
  context/
    manager.py         # ContextManager, CompactEvent
    strategies.py      # CompactStrategy, NoCompact, TieredCompact, SlidingWindowCompact
    hardware.py        # HardwareProfile, detect_hardware()
  prompts/
    templates.py       # Tool prompt builders (prompt-injected path)
    nudges.py          # Retry and step-enforcement nudge templates
  tools/
    respond.py         # Synthetic respond tool (respond_tool(), respond_spec())
  proxy/
    __main__.py        # CLI entry point: python -m forge.proxy
    proxy.py           # ProxyServer — programmatic start/stop API
    server.py          # Raw asyncio HTTP server, SSE streaming
    handler.py         # Request handler — bridge between HTTP and run_inference
    convert.py         # OpenAI messages ↔ forge Messages conversion
tests/
  unit/                # Deterministic tests — no LLM backend required
  eval/                # Eval harness — model qualification against real backends

• Forge Proxy Installation — Standalone platform installation, profiles, updates, recovery, and uninstall
• User Guide — Usage patterns, multi-turn, context management, guardrails, slot worker, long-running session advisory
• Model Guide — Which model and backend for your hardware
• Backend Setup — Backend installation and server setup
• Eval Guide — Eval harness CLI reference, batch eval
• Architecture — Full design document
• Workflow Internals — Workflow design and runner internals
• Contributing — How to set up, test, and add new backends or scenarios

The forge guardrail framework and ablation study are published as:

Zambelli, A. Forge: Closing the Agentic Reliability Gap Between Self-Hosted and Frontier Language Models. https://doi.org/10.1145/3786335.3813193

A pre-publication preprint is also available at docs/forge_ieee_preprint.pdf — kept as a historical artifact. Cite the published version above; the DOI link may not resolve immediately depending on the publisher's release timing.

MIT — Copyright (c) 2025-2026 Antoine Zambelli