mirror of
https://github.com/coder/coder.git
synced 2026-10-16 14:48:56 +08:00
reordered sections and removed hallucinations
This commit is contained in:
@@ -1,4 +1,4 @@
|
||||
AI Bridge Proxy
|
||||
# AI Bridge Proxy
|
||||
|
||||
AI Bridge Proxy extends [AI Bridge](../index.md) to support clients that don't allow base URL overrides.
|
||||
While AI Bridge requires clients to support custom base URLs, many popular AI coding tools lack this capability.
|
||||
|
||||
@@ -134,7 +134,6 @@ To configure a specific trace endpoint:
|
||||
|
||||
```bash
|
||||
export CODER_TRACE_ENABLE=true
|
||||
export CODER_TRACE_ENDPOINT=<your-otel-collector-endpoint>
|
||||
```
|
||||
|
||||
### What is Traced
|
||||
@@ -190,16 +189,6 @@ rates — all labeled by user, model, and provider.
|
||||
For general Prometheus setup, see the
|
||||
[Coder Prometheus documentation](https://coder.com/docs/admin/integrations/prometheus).
|
||||
|
||||
## Rate limiting
|
||||
|
||||
AI Bridge supports rate limiting to protect upstream providers and your
|
||||
deployment:
|
||||
|
||||
| Setting | Description | Default |
|
||||
|---------|-------------|---------|
|
||||
| `CODER_AIBRIDGE_MAX_RPS` | Maximum requests per second per replica. Set to `0` to disable. | `0` |
|
||||
| `CODER_AIBRIDGE_MAX_CONCURRENT` | Maximum concurrent requests per replica. Set to `0` to disable. | `0` |
|
||||
|
||||
## Next steps
|
||||
|
||||
- [Setup](./setup.md) — Enable AI Bridge and configure providers.
|
||||
|
||||
@@ -5,6 +5,25 @@
|
||||
AI Bridge runs as an in-process component within `coderd` (the Coder control
|
||||
plane). It does not require a separate deployment or infrastructure.
|
||||
|
||||
### Key architectural points
|
||||
|
||||
- **Developers never hold provider API keys.** They authenticate with their
|
||||
existing Coder credentials.
|
||||
- **The enterprise holds one key per provider.** These are configured centrally
|
||||
in `coderd` via environment variables or Helm values.
|
||||
- **Template-level configuration.** Configure AI Bridge once in a workspace
|
||||
template, and it propagates to every workspace created from that template.
|
||||
- **Universal client compatibility.** Works with any AI tool that supports base
|
||||
URL overrides, including Claude Code, Codex, Cline, Zed, and more. For tools
|
||||
that don't support base URL overrides, [AI Bridge Proxy](./ai-bridge-proxy/index.md)
|
||||
can intercept traffic transparently.
|
||||
|
||||
## Implementation Details
|
||||
|
||||
`coderd` runs an in-memory instance of `aibridged`, whose logic is mostly contained in https://github.com/coder/aibridge. In future releases we will support running external instances for higher throughput and complete memory isolation from `coderd`.
|
||||
|
||||

|
||||
|
||||
### Request lifecycle
|
||||
|
||||
1. A template admin configures AI Bridge base URLs and session tokens in the
|
||||
@@ -25,25 +44,6 @@ plane). It does not require a separate deployment or infrastructure.
|
||||
9. Bridge logs usage (tokens, tools, model, user) and returns the final response
|
||||
to the agent.
|
||||
|
||||
### Key architectural points
|
||||
|
||||
- **Developers never hold provider API keys.** They authenticate with their
|
||||
existing Coder credentials.
|
||||
- **The enterprise holds one key per provider.** These are configured centrally
|
||||
in `coderd` via environment variables or Helm values.
|
||||
- **Template-level configuration.** Configure AI Bridge once in a workspace
|
||||
template, and it propagates to every workspace created from that template.
|
||||
- **Universal client compatibility.** Works with any AI tool that supports base
|
||||
URL overrides, including Claude Code, Codex, Cline, Zed, and more. For tools
|
||||
that don't support base URL overrides, [AI Bridge Proxy](./ai-bridge-proxy/index.md)
|
||||
can intercept traffic transparently.
|
||||
|
||||
## Implementation Details
|
||||
|
||||
`coderd` runs an in-memory instance of `aibridged`, whose logic is mostly contained in https://github.com/coder/aibridge. In future releases we will support running external instances for higher throughput and complete memory isolation from `coderd`.
|
||||
|
||||

|
||||
|
||||
## Supported APIs
|
||||
|
||||
API support is broken down into two categories:
|
||||
|
||||
Reference in New Issue
Block a user