74bbd8e6d2
- Parse tools/tool_choice from OpenAI and Anthropic requests - Inject tool definitions into system prompt via toolemulation - Parse action blocks (```json action) from model responses - Retry logic for forced tool_choice (any/required) - Return proper tool_calls / tool_use in responses - Support streaming tools via collect-and-replay pattern - Add tool history projection (assistant tool_calls + tool results) - Model ID normalization: use official names (Qwen3.6-Plus, etc.) - Fix resolveSessionMode to use Fresh mode when tools present
345 lines
6.9 KiB
Markdown
345 lines
6.9 KiB
Markdown
# lingma-ipc-proxy
|
||
|
||
[English](./README.md) | [简体中文](./README.zh-CN.md)
|
||
|
||
A standalone Go backend that talks to Lingma over Lingma's local pipe or websocket transport and exposes:
|
||
|
||
- `GET /v1/models`
|
||
- `POST /v1/messages`
|
||
- `POST /v1/chat/completions`
|
||
|
||
Current scope:
|
||
|
||
- supports both non-streaming and streaming responses
|
||
- one request at a time
|
||
- supports Windows named-pipe transport and local websocket transport
|
||
- directly uses Lingma IPC, not DOM/CDP
|
||
- OpenAI-compatible `tools` / `tool_choice` support (tool emulation via prompt engineering)
|
||
- Anthropic-compatible `tools` / `tool_choice` support
|
||
|
||
## Run
|
||
|
||
```powershell
|
||
cd C:\Workspace\Personal\lingma-ipc-proxy
|
||
go run .\cmd\lingma-ipc-proxy
|
||
```
|
||
|
||
## Config File
|
||
|
||
The proxy can load a JSON config file so you do not need to carry a long command line every time.
|
||
|
||
Default lookup:
|
||
|
||
```text
|
||
./lingma-ipc-proxy.json
|
||
```
|
||
|
||
You can also point to an explicit file:
|
||
|
||
```powershell
|
||
.\dist\lingma-ipc-proxy.exe --config .\config.example.json
|
||
```
|
||
|
||
Resolution order:
|
||
|
||
- built-in defaults
|
||
- JSON config file
|
||
- environment variables
|
||
- command-line flags
|
||
|
||
An example config is included at:
|
||
|
||
- `config.example.json`
|
||
|
||
A practical setup is to copy it to `lingma-ipc-proxy.json`, adjust the values once, and then start the proxy without a long flag list.
|
||
|
||
Recommended layout:
|
||
|
||
```json
|
||
{
|
||
"host": "127.0.0.1",
|
||
"port": 8095,
|
||
"transport": "auto",
|
||
"mode": "chat",
|
||
"session_mode": "reuse",
|
||
"timeout": 120,
|
||
"cwd": "C:/Workspace/Personal/lingma-ipc-proxy",
|
||
"shell_type": "powershell",
|
||
"current_file_path": "",
|
||
"pipe": "",
|
||
"websocket_url": ""
|
||
}
|
||
```
|
||
|
||
## macOS / Linux
|
||
|
||
This project also works on macOS and Linux via **WebSocket transport**. The Windows named-pipe transport is automatically skipped on non-Windows platforms.
|
||
|
||
### Run on macOS
|
||
|
||
```bash
|
||
cd ~/OpenSources/lingma-ipc-proxy
|
||
go run ./cmd/lingma-ipc-proxy --transport websocket --port 8095
|
||
|
||
# Or use auto-detect (will discover websocket port from Lingma's shared client cache)
|
||
go run ./cmd/lingma-ipc-proxy --port 8095
|
||
```
|
||
|
||
### Build on macOS / Linux
|
||
|
||
```bash
|
||
cd ~/OpenSources/lingma-ipc-proxy
|
||
go build -o ./dist/lingma-ipc-proxy ./cmd/lingma-ipc-proxy
|
||
```
|
||
|
||
### macOS Config Example
|
||
|
||
```json
|
||
{
|
||
"host": "127.0.0.1",
|
||
"port": 8095,
|
||
"transport": "websocket",
|
||
"mode": "agent",
|
||
"shell_type": "zsh",
|
||
"session_mode": "auto",
|
||
"timeout": 120
|
||
}
|
||
```
|
||
|
||
## Build
|
||
|
||
Build a Windows executable:
|
||
|
||
```powershell
|
||
cd C:\Workspace\Personal\lingma-ipc-proxy
|
||
.\scripts\build.ps1
|
||
```
|
||
|
||
Default output:
|
||
|
||
```text
|
||
dist\lingma-ipc-proxy.exe
|
||
```
|
||
|
||
## Release
|
||
|
||
GitHub Actions can publish a GitHub Release automatically.
|
||
|
||
Trigger rules:
|
||
|
||
- push a tag matching `v*`, for example `v0.1.0`
|
||
- or run the `Release` workflow manually and pass a tag
|
||
|
||
Example:
|
||
|
||
```powershell
|
||
git tag v0.1.0
|
||
git push origin v0.1.0
|
||
```
|
||
|
||
Release assets:
|
||
|
||
- `lingma-ipc-proxy_<tag>_windows_amd64.exe`
|
||
- `lingma-ipc-proxy_<tag>_windows_amd64.zip`
|
||
- `lingma-ipc-proxy_<tag>_sha256.txt`
|
||
|
||
Direct Go build command:
|
||
|
||
```powershell
|
||
$env:CGO_ENABLED = "0"
|
||
$env:GOOS = "windows"
|
||
$env:GOARCH = "amd64"
|
||
go build -trimpath -ldflags "-s -w" -o .\dist\lingma-ipc-proxy.exe .\cmd\lingma-ipc-proxy
|
||
```
|
||
|
||
Run the built binary:
|
||
|
||
```powershell
|
||
.\dist\lingma-ipc-proxy.exe --host 127.0.0.1 --port 8095 --session-mode auto
|
||
.\dist\lingma-ipc-proxy.exe --transport websocket --ws-url ws://127.0.0.1:36510 --port 8095
|
||
```
|
||
|
||
## Windows Service
|
||
|
||
For this project, the correct deployment shape is a native local process, not Docker. The proxy talks to Lingma over local pipe or websocket transport, so it should run on the same host as Lingma itself.
|
||
|
||
### NSSM
|
||
|
||
Build first:
|
||
|
||
```powershell
|
||
.\scripts\build.ps1
|
||
```
|
||
|
||
Install with NSSM:
|
||
|
||
```powershell
|
||
.\scripts\install-nssm-service.ps1 -NssmPath C:\Tools\nssm\nssm.exe
|
||
```
|
||
|
||
This wraps:
|
||
|
||
```powershell
|
||
nssm.exe install LingmaIpcProxy C:\Workspace\Personal\lingma-ipc-proxy\dist\lingma-ipc-proxy.exe --host 127.0.0.1 --port 8095 --session-mode auto
|
||
nssm.exe set LingmaIpcProxy AppDirectory C:\Workspace\Personal\lingma-ipc-proxy
|
||
nssm.exe start LingmaIpcProxy
|
||
```
|
||
|
||
### WinSW
|
||
|
||
Prepare the executable:
|
||
|
||
```powershell
|
||
.\scripts\build.ps1
|
||
```
|
||
|
||
Put a WinSW binary at:
|
||
|
||
```text
|
||
dist\WinSW-x64.exe
|
||
```
|
||
|
||
Then generate the wrapper files:
|
||
|
||
```powershell
|
||
.\scripts\install-winsw-service.ps1
|
||
```
|
||
|
||
That script creates:
|
||
|
||
- `LingmaIpcProxy.exe`
|
||
- `LingmaIpcProxy.xml`
|
||
|
||
Then install/start:
|
||
|
||
```powershell
|
||
.\LingmaIpcProxy.exe install
|
||
.\LingmaIpcProxy.exe start
|
||
```
|
||
|
||
The WinSW XML template lives at:
|
||
|
||
- `scripts\lingma-ipc-proxy.xml.template`
|
||
|
||
## Flags
|
||
|
||
```powershell
|
||
go run .\cmd\lingma-ipc-proxy --port 8095 --session-mode auto
|
||
```
|
||
|
||
- `--host`
|
||
- `--port`
|
||
- `--transport`
|
||
- `--pipe`
|
||
- `--ws-url`
|
||
- `--cwd`
|
||
- `--current-file-path`
|
||
- `--mode`
|
||
- `--shell-type`
|
||
- `--session-mode`
|
||
- `reuse`: keep using the sticky Lingma session
|
||
- `fresh`: create a temporary session for the request and delete it after completion
|
||
- `auto`: single-turn requests reuse; requests with system/history use a temporary fresh session and delete it after completion
|
||
- `--timeout`
|
||
|
||
## Environment
|
||
|
||
- `LINGMA_PROXY_TRANSPORT`
|
||
- `LINGMA_IPC_PIPE`
|
||
- `LINGMA_PROXY_WS_URL`
|
||
- `LINGMA_PROXY_HOST`
|
||
- `LINGMA_PROXY_PORT`
|
||
- `LINGMA_PROXY_CWD`
|
||
- `LINGMA_PROXY_CURRENT_FILE_PATH`
|
||
- `LINGMA_PROXY_MODE`
|
||
- `LINGMA_PROXY_SHELL_TYPE`
|
||
- `LINGMA_PROXY_SESSION_MODE`
|
||
- `LINGMA_PROXY_TIMEOUT_SECONDS`
|
||
|
||
## Examples
|
||
|
||
Anthropic non-streaming:
|
||
|
||
```powershell
|
||
$body = @{
|
||
model = "dashscope_qwen3_coder"
|
||
messages = @(
|
||
@{ role = "user"; content = "请只回复：ANTHROPIC_OK" }
|
||
)
|
||
stream = $false
|
||
} | ConvertTo-Json -Depth 8
|
||
|
||
Invoke-RestMethod `
|
||
-Method Post `
|
||
-Uri http://127.0.0.1:8095/v1/messages `
|
||
-ContentType "application/json" `
|
||
-Body $body
|
||
```
|
||
|
||
Anthropic streaming:
|
||
|
||
```powershell
|
||
$body = @{
|
||
model = "dashscope_qwen3_coder"
|
||
messages = @(
|
||
@{ role = "user"; content = "请只回复：ANTHROPIC_STREAM_OK" }
|
||
)
|
||
stream = $true
|
||
} | ConvertTo-Json -Depth 8
|
||
|
||
curl.exe -N `
|
||
-H "Content-Type: application/json" `
|
||
-d $body `
|
||
http://127.0.0.1:8095/v1/messages
|
||
```
|
||
|
||
OpenAI non-streaming:
|
||
|
||
```powershell
|
||
$body = @{
|
||
model = "dashscope_qwen3_coder"
|
||
messages = @(
|
||
@{ role = "user"; content = "请只回复：OPENAI_OK" }
|
||
)
|
||
stream = $false
|
||
} | ConvertTo-Json -Depth 8
|
||
|
||
Invoke-RestMethod `
|
||
-Method Post `
|
||
-Uri http://127.0.0.1:8095/v1/chat/completions `
|
||
-ContentType "application/json" `
|
||
-Body $body
|
||
```
|
||
|
||
OpenAI streaming:
|
||
|
||
```powershell
|
||
$body = @{
|
||
model = "dashscope_qwen3_coder"
|
||
messages = @(
|
||
@{ role = "user"; content = "请只回复：OPENAI_STREAM_OK" }
|
||
)
|
||
stream = $true
|
||
} | ConvertTo-Json -Depth 8
|
||
|
||
curl.exe -N `
|
||
-H "Content-Type: application/json" `
|
||
-d $body `
|
||
http://127.0.0.1:8095/v1/chat/completions
|
||
```
|
||
|
||
## Streaming shape
|
||
|
||
Anthropic streaming emits SSE events compatible with the `messages` API shape:
|
||
|
||
- `message_start`
|
||
- `content_block_start`
|
||
- `content_block_delta`
|
||
- `content_block_stop`
|
||
- `message_delta`
|
||
- `message_stop`
|
||
|
||
OpenAI streaming emits `chat.completion.chunk` payloads as `data:` lines and ends with:
|
||
|
||
- `data: [DONE]`
|