kimi.http

# Kimi / Moonshot API connectivity tests
# Run with VS Code "REST Client" extension or JetBrains HTTP Client.
# These mirror the exact calls used to diagnose the opencode "Unauthorized" error.

# Secrets (moonshotKey, kimiKey, moonshotGlobalKey, piToken) live in
# http-client.private.env.json at the project root (git-ignored). JetBrains
# HTTP Client resolves {{vars}} from there. For VS Code REST Client, define
# them in settings.json under "rest-client.environmentVariables" instead.
#
# Keys are generated at https://platform.moonshot.cn (valid on api.moonshot.cn
# ONLY - rejected by api.kimi.com/coding and api.moonshot.ai).
# TODO: kimiKey -> generate a real Kimi Code key at https://www.kimi.com/code
# TODO: moonshotGlobalKey -> only if you have a https://platform.moonshot.ai (global) account

###

# 1. Kimi For Coding endpoint - what opencode calls for kimi-for-coding/k3
#    Expected with current key: 401 {"type":"invalid_authentication_error"}
#    Expected with a real Kimi Code key: 200 with message content
POST https://api.kimi.com/coding/v1/messages
Authorization: Bearer {{kimiKey}}
x-api-key: {{kimiKey}}
anthropic-version: 2023-06-01
Content-Type: application/json

{
  "model": "k3",
  "max_tokens": 20,
  "messages": [{ "role": "user", "content": "say ok" }]
}

###

# 2. Moonshot CN platform - list models (cheapest key-validity check)
#    Expected with current key: 200 + model list (key IS valid here)
GET https://api.moonshot.cn/v1/models
Authorization: Bearer {{moonshotKey}}

###

# 3. Moonshot CN platform - call kimi-k3 (Anthropic-compatible API)
#    Expected with current key: 429 exceeded_current_quota_error
#    ("account suspended due to insufficient balance") -> key valid, no credit.
#    After recharging: 200.
POST https://api.moonshot.cn/anthropic/v1/messages
x-api-key: {{moonshotKey}}
anthropic-version: 2023-06-01
Content-Type: application/json

{
  "model": "kimi-k3",
  "max_tokens": 20,
  "messages": [{ "role": "user", "content": "say ok" }]
}

###

# 4. Moonshot CN platform - chat completions (OpenAI-compatible API)
#    Same expectation as #3: 429 until balance is topped up.
POST https://api.moonshot.cn/v1/chat/completions
Authorization: Bearer {{moonshotKey}}
Content-Type: application/json

{
  "model": "kimi-k3",
  "max_tokens": 20,
  "messages": [{ "role": "user", "content": "say ok" }]
}

###

# 5. Moonshot GLOBAL platform - list models
#    Expected with current key: 401 (CN platform key is not valid globally)
GET https://api.moonshot.ai/v1/models
Authorization: Bearer {{moonshotGlobalKey}}

###

# 6. EXACTLY what pi (this agent) uses - OAuth subscription token
#    Endpoint : https://api.kimi.com/coding/v1/messages  (Anthropic Messages API)
#    Auth     : Authorization: Bearer <OAuth access token>  (NOT an sk- API key)
#    Token exp: 2026-08-27 23:28:16 UTC -> VALID at time of writing
#    NOTE: token lives ~15 min; pi auto-refreshes it via the refresh token in
#    ~/.pi/agent/auth.json. If you get 401, re-copy a fresh token from that file
#    into http-client.private.env.json.
#    Expected while valid: 200. This is the call that proves why pi works.

POST https://api.kimi.com/coding/v1/messages
Authorization: Bearer {{piToken}}
anthropic-version: 2023-06-01
Content-Type: application/json

{
  "model": "k3",
  "max_tokens": 20,
  "messages": [{ "role": "user", "content": "say ok" }]
}