bios-mcp 0.1.1-dev.22 → 0.1.1-dev.24
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/README.md +11 -4
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -42,8 +42,9 @@ not claimed without an explicit acknowledgement from the endpoint.
|
|
|
42
42
|
|
|
43
43
|
1. `get_inference_gpu_options` returns inference-fit configurations joined to
|
|
44
44
|
authoritative deployment prices and stock. Prefer the MODEL-ADDRESSED form
|
|
45
|
-
(`model=<
|
|
46
|
-
server resolves the model facts (vision-aware)
|
|
45
|
+
(`model=<catalog_model_id>` — an id from `list_models`/`search_models`,
|
|
46
|
+
optionally `revision`): the server resolves the model facts (vision-aware)
|
|
47
|
+
and answers with computed
|
|
47
48
|
`min_gpus`, `valid_counts`, and `bookable_counts` — the exact minimums the
|
|
48
49
|
create gate enforces, so client-supplied facts can never understate a
|
|
49
50
|
minimum. `unknown` is not treated as available; never offer a count below
|
|
@@ -108,8 +109,14 @@ A `503 ADMISSION_UNAVAILABLE` is a transient market outage, never a capacity
|
|
|
108
109
|
verdict — retry shortly, never conclude out-of-stock from it. See
|
|
109
110
|
`get_platform_guide` topic `capacity_errors` for the recovery playbook.
|
|
110
111
|
|
|
111
|
-
|
|
112
|
-
|
|
112
|
+
Models are selectable only from the platform's own catalog: `list_models`
|
|
113
|
+
and `search_models` read the hosted registry, and `create_training_job`,
|
|
114
|
+
`preflight_training_job`, `create_inference`, and `preflight_inference`
|
|
115
|
+
validate the model id against it, answering `MODEL_NOT_HOSTED` — never a
|
|
116
|
+
live Hugging Face lookup. Catalog models are pre-mirrored, so raw access
|
|
117
|
+
tokens are intentionally excluded from MCP tool inputs and no integration is
|
|
118
|
+
needed for models; Hugging Face integrations remain for dataset imports via
|
|
119
|
+
`import_huggingface_dataset`.
|
|
113
120
|
|
|
114
121
|
`create_training_job` and `resume_training_job` accept an optional
|
|
115
122
|
`idempotency_key`. Reuse the same value after a timeout so the API replays the
|