qflash/README.md

# Qflash: OpenAI-compatible gateway for Qwen3.8 models

## About

Qflash is a standalone, single-binary gateway that exposes Qwen3.8 model spaces on Hugging Face through a standard OpenAI-compatible API. It automatically detects and supports multiple upstream protocols:

- gradio `/respond` endpoints (such as `https://microhero-qwen3-8-27b-uncensored-chat.hf.space`, running live on ZeroGPU)
- direct OpenAI `/v1/chat/completions` endpoints (such as `https://wanyamaelis-qwen3-8-27b.hf.space` and `https://apathy-exe-qwen3-8-flash-next.hf.space`, running `llama-server` on CPU)
- gradio `/chat_response` endpoints (such as `https://halvo78-qwen3-8-flash-next-playground.hf.space`)

It translates standard `/v1/chat/completions` and `/v1/models` requests and Server-Sent Events (SSE) stream protocols into upstream formats, allowing any standard OpenAI-compatible client, agent, or IDE to interface with Qwen3.8 models without modification.

## Features

- openai-compatible chat completions (streaming and non-streaming)
- automatic upstream failover across ZeroGPU and non-ZeroGPU endpoints
- smart endpoint cooldown (5 minutes on quota exhaustion, 30 seconds on network errors)
- automatic upstream endpoint detection (`respond`, `openai`, `chat_response`)
- deep reasoning extraction with thinking trace passthrough (`<think>` tags and blockquotes mapped to `reasoning_content`)
- stateful streaming tool call interception (`StreamToolCallFilter`) with zero XML or JSON leakage into `delta.content`
- real-time token streaming with incremental SSE delivery
- support for instruct mode via standard `reasoning_effort: "none"`
- zero-dependency SOCKS5 proxy client (RFC 1928 / RFC 1929) with domain resolution (`socks5h://`), IPv4, IPv6, and auth
- bring your own key (BYOK) pass-through support via `Authorization: Bearer` or CLI flags
- zero-gpu quota authentication via free Hugging Face personal tokens (`HF_TOKEN`)
- fibonacci backoff retry on transient upstream errors
- zero external dependencies (Go standard library only)

## Supported upstream spaces

| Space | Model | Hardware | Protocol | Notes |
|---|---|---|---|---|
| `MicroHERO/qwen3.8-27b-uncensored-chat` *(primary)* | Qwen3.8-27B Uncensored | ZeroGPU (A10G) | Gradio `/respond` | Fast live GPU inference (~2s), uncensored weights |
| `Wanyamaelis/Qwen3.8-27B` *(fallback)* | Qwen3.8-27B (MTP) | CPU basic (8 vCPU) | Native OpenAI `/v1` | Non-ZeroGPU, no quota limits, fast speculative decoding (~5s) |
| `apathy-exe/Qwen3.8-27B` | Qwen3.8-27B (MTP) | CPU (OpenMP/AVX512) | Native OpenAI `/v1` | Non-ZeroGPU, no quota limits, speculative decoding |
| `apathy-exe/Qwen3.8-Flash-Next` | Qwen3.8-Flash-Next (~177B) | CPU (OpenMP/AVX512) | Native OpenAI `/v1` | Non-ZeroGPU, no quota limits, full 177B Flash-Next model |
| `Halvo78/qwen3-8-flash-next-playground` | Qwen3.8-Flash-Next | CPU basic | Gradio `/chat_response` | Sandbox client; requires BYOK API key/base URL |

### Auto-failover and ZeroGPU quota handling

Spaces running on ZeroGPU provide free A10G compute, but anonymous requests share a small pool per IP address. When ZeroGPU runs limit is reached, upstream returns a quota error (`429` or `ZeroGPU runs limit`).

Qflash handles this seamlessly:

- with auto-failover enabled (default), when a ZeroGPU space hits its runs limit, it is placed on a 5-minute cooldown and the gateway automatically fails over to the next configured endpoint (e.g. `wanyamaelis-qwen3-8-27b`, which runs on CPU with zero quota limits)
- transient network errors trigger a shorter 30-second cooldown before retrying
- you can also provide a free Hugging Face personal access token (`https://huggingface.co/settings/tokens`) via the `HF_TOKEN` environment variable, the `-hf-token` CLI flag, or the `Authorization: Bearer hf_...` header to authenticate ZeroGPU requests directly

## Installation

You need Go 1.22 or newer (tested on Go 1.26). Install the latest release straight from the Git repository with `go install`:

```bash
go install code.luxferre.top/luxferre/qflash@latest
```

This fetches the module from `https://code.luxferre.top/luxferre/qflash.git` and places the `qflash` binary in `$(go env GOPATH)/bin`. Make sure that directory is on your `PATH`.

*(Note: If installing right after a new commit has been pushed, bypass any proxy cache with `GOPROXY=direct go install code.luxferre.top/luxferre/qflash@latest`).*

If you prefer to build from a local checkout instead:

```bash
git clone https://code.luxferre.top/luxferre/qflash.git
cd qflash
go install .
```

Alternatively, build directly from source using `make`:

```bash
make qflash
```

This produces the `bin/qflash` binary for your platform.

## Models served

The gateway advertises the following models under `/v1/models`:

| Model ID | Target model | Description |
|---|---|---|
| `Qwen/Qwen3.8-27B-Uncensored` | `Qwen/Qwen3.8-27B-Uncensored` | Default primary live model |
| `Qwen/Qwen3.8-Flash-Next` | `Qwen/Qwen3.8-Flash-Next` | Flash-Next model identifier |
| `qwen3.8-27b-uncensored` | `Qwen/Qwen3.8-27B-Uncensored` | Standard lowercase alias |
| `qwen3.8-flash-next` | `Qwen/Qwen3.8-Flash-Next` | Flash-Next lowercase alias |
| `qwen-flash-next` | `Qwen/Qwen3.8-Flash-Next` | Shorthand alias |
| `qwen-flash` | `Qwen/Qwen3.8-Flash-Next` | Quick convenience alias |
| `qwen` | Default model | Generic shorthand alias |

Any unlisted custom model name requested by the client is passed through directly.

## Usage

Run the gateway with default auto-failover endpoints:

```bash
qflash
```

By default, this listens on `http://127.0.0.1:8080` with failover configured across `MicroHERO` (ZeroGPU), `wanyamaelis` (CPU non-ZeroGPU), and `apathy-exe` endpoints.

Available flags:

- `-port` — tcp port to listen on (default `8080`)
- `-endpoints` / `-endpoint` — comma-separated upstream space or OpenAI URLs (default list of 4 endpoints, `QFLASH_ENDPOINTS` / `QFLASH_ENDPOINT` env)
- `-failover` / `-auto-failover` — enable automatic failover across endpoints on quota exhaustion or error (default `true`, `QFLASH_FAILOVER` env)
- `-mode` — upstream protocol mode: `auto`, `respond`, `chat_response`, `openai` (default `auto`, `QFLASH_MODE` env)
- `-model` — exposed model name (default `Qwen/Qwen3.8-27B-Uncensored`, `QFLASH_MODEL` env)
- `-thinking` / `-enable-thinking` — enable chain-of-thought reasoning by default (default `true`)
- `-hf-token` / `-token` — Hugging Face API token for ZeroGPU quota or private spaces (`HF_TOKEN` env)
- `-api-key` — upstream inference engine API key for BYOK mode (`OPENAI_API_KEY` / `QWEN_API_KEY` env)
- `-base-url` — upstream inference engine base URL for BYOK mode (`OPENAI_BASE_URL` / `QWEN_BASE_URL` env)
- `-socks` / `-proxy` / `-socks5` — SOCKS5 proxy URL, e.g. `socks5://127.0.0.1:1080` (`ALL_PROXY` env)
- `-user-agent` / `-ua` — custom User-Agent sent to upstream

Endpoints served:

- `GET  /models` and `GET  /v1/models`
- `POST /chat/completions` and `POST /v1/chat/completions`

Example request with curl:

```bash
curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen","messages":[{"role":"user","content":"Explain quantum superposition in one sentence."}],"stream":false}'
```

Streaming example:

```bash
curl -N http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen","messages":[{"role":"user","content":"Count from 1 to 5."}],"stream":true}'
```

Instruct mode example (disables thinking):

```bash
curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen","messages":[{"role":"user","content":"Hello!"}],"reasoning_effort":"none"}'
```

Using custom upstream spaces or a single endpoint:

```bash
# Point to a single CPU llama-server without failover
qflash -endpoint https://wanyamaelis-qwen3-8-27b.hf.space -failover=false

# Point to Halvo78 playground with your own API credentials
qflash -endpoint https://halvo78-qwen3-8-flash-next-playground.hf.space -api-key "sk-..." -base-url "https://api.openai.com/v1"
```

Python OpenAI SDK integration:

```python
from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8080/v1",
    api_key="sk-dummy"
)

stream = client.chat.completions.create(
    model="qwen",
    messages=[
        {"role": "user", "content": "Prove that the sum of the first n odd numbers is n^2."}
    ],
    stream=True
)

for chunk in stream:
    delta = chunk.choices[0].delta
    if hasattr(delta, "reasoning_content") and delta.reasoning_content:
        print(f"[THINK] {delta.reasoning_content}", end="", flush=True)
    if delta.content:
        print(delta.content, end="", flush=True)
print()
```

## Credits

Created by Luxferre in 2026, released into the public domain with no warranties.