Claude 014dcd4698 Pair each Kimi wire shape with a model its endpoint serves ...

Moonshot's Anthropic-compatible /anthropic surface only serves the
current coding flagship: requesting kimi-k2-thinking there returns
resource_not_found_error ('Not found the model kimi-k2-thinking or
Permission denied'), while kimi-k3 works — it's the model Moonshot's
own Claude Code guide configures. Swap the pairing: kimi-k3 rides the
Anthropic shape and kimi-k2-thinking the OpenAI shape (which serves
every platform model). Model strings stay distinct across the two
provider records, keeping routing unambiguous.

2026-07-21 12:14:15 +00:00

..

bootstrap_test.go

[management,proxy,client] 0.74.0 version (#6563)

2026-07-01 12:45:14 +02:00

chat_test.go

Pair each Kimi wire shape with a model its endpoint serves

2026-07-21 12:14:15 +00:00

guardrail_test.go

Use the full Bedrock inference-profile id in the e2e matrix

2026-07-21 11:49:42 +00:00

main_test.go

[management,proxy,client] 0.74.0 version (#6563)

2026-07-01 12:45:14 +02:00

management_test.go

[management,proxy,client] 0.74.0 version (#6563)

2026-07-01 12:45:14 +02:00

skiptls_test.go

[management,proxy] Add per-provider skip_tls_verification for agent-network (#6630)

2026-07-01 20:43:15 +02:00

vllm_test.go

[management] Add vLLM e2e test (#6649)

2026-07-02 15:36:51 +02:00