llm-relay 0.68.0 → 0.68.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/cli.d.ts +14 -0
- package/dist/cli.js +105 -5
- package/dist/cli.js.map +1 -1
- package/dist/hedge-trigger.js +3 -2
- package/dist/hedge-trigger.js.map +1 -1
- package/dist/mcp/lane-runner.d.ts +19 -1
- package/dist/mcp/lane-runner.js +78 -2
- package/dist/mcp/lane-runner.js.map +1 -1
- package/dist/mcp/server.d.ts +3 -1
- package/dist/mcp/server.js +25 -3
- package/dist/mcp/server.js.map +1 -1
- package/dist/server.js +1 -1
- package/dist/server.js.map +1 -1
- package/package.json +1 -1
- package/skills/llm-relay/SKILL.md +9 -1
package/package.json
CHANGED
|
@@ -211,6 +211,14 @@ Keep the parent on its normal Codex provider and define a named child agent unde
|
|
|
211
211
|
local Codex clients: the parent retains native Codex orchestration, while the child spends the
|
|
212
212
|
configured provider pool. The `llm-relay` profile is an all-relay mode and routes the parent too.
|
|
213
213
|
|
|
214
|
+
⚠ Codex desktop collaboration has a host-side limitation measured on Codex 0.151.0: with a
|
|
215
|
+
ChatGPT account, the collaboration launcher validates a child model against the parent account
|
|
216
|
+
before contacting llm-relay and ignores the child's `model_provider`. A `pool/medium` child can
|
|
217
|
+
therefore fail with HTTP 400 (`model is not supported when using Codex with a ChatGPT account`).
|
|
218
|
+
In the app, use the `llm-relay` MCP `dispatch` tool for relay-backed work instead. The generated
|
|
219
|
+
agent files remain useful for Codex clients that honor custom providers (and are still provisioned
|
|
220
|
+
by the installer).
|
|
221
|
+
|
|
214
222
|
A global npm install provisions the `llm-relay` Responses provider in `~/.codex/config.toml` and
|
|
215
223
|
creates the relay-backed `default` and `relay_coding` agents under `~/.codex/agents/` when they are
|
|
216
224
|
absent. It preserves existing Codex config and agent files; use the manual snippets below if the
|
|
@@ -532,7 +540,7 @@ llm-relay models -p nim # live roster per provider (listed ≠ servable —
|
|
|
532
540
|
llm-relay keys # check EVERY configured credential slot
|
|
533
541
|
llm-relay pools --probe # one completion per unique deployment, via one serviceable slot
|
|
534
542
|
llm-relay ping # latency/stability probe across providers
|
|
535
|
-
llm-relay telemetry # JSON
|
|
543
|
+
llm-relay telemetry # provider-health/runtime-observation JSON (not dispatch quota JSON)
|
|
536
544
|
```
|
|
537
545
|
|
|
538
546
|
Runtime endpoints on the running proxy: `/registry`, `/candidates`, `/offload?client=<name>` (GET/POST),
|