mirror of
https://github.com/mksglu/context-mode.git
synced 2026-10-02 04:14:38 +08:00
refactor(pricing): single catalog under src/session; structured cost event
Consolidates the 5 vendor price files into one src/session/model-prices.json (61 models), moves the module to src/session/pricing.ts beside its only consumer (extract.ts), and drops the dev-only litellm NOTES from tracking (litellm base stays gitignored). Wave 2b: extractAgentUsage emits model_id + input/output/cache_read/cache_creation tokens + cost_usd as structured top-level event fields, replacing the colon-string, so the platform persists them as columns. 45 tests green, tsc clean.
This commit is contained in:
Mert Koseoglu
+1
-2
@@ -40,8 +40,7 @@ context-mode-guidance-*/
|
||||
.vibetree/
|
||||
.cw/
|
||||
# Dev-only pricing source: the full litellm catalog (~1.5MB) is the manual
|
||||
# refresh base for src/pricing/sources/*.json — never bundled into the plugin.
|
||||
# Refresh: see tools/pricing/litellm-NOTES.md. Keep NOTES tracked, ignore the blob.
|
||||
# refresh base for src/session/model-prices.json — never bundled into the plugin.
|
||||
tools/pricing/litellm-catalog.json
|
||||
/.cocoindex_code/
|
||||
/.kilo/
|
||||
|
||||
@@ -1,82 +0,0 @@
|
||||
{
|
||||
"claude-opus-4-8": {
|
||||
"input_per_mtok": 5.00,
|
||||
"output_per_mtok": 25.00,
|
||||
"cache_read_per_mtok": 0.50,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-opus-4-7": {
|
||||
"input_per_mtok": 5.00,
|
||||
"output_per_mtok": 25.00,
|
||||
"cache_read_per_mtok": 0.50,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-opus-4-6": {
|
||||
"input_per_mtok": 5.00,
|
||||
"output_per_mtok": 25.00,
|
||||
"cache_read_per_mtok": 0.50,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-opus-4-5": {
|
||||
"input_per_mtok": 5.00,
|
||||
"output_per_mtok": 25.00,
|
||||
"cache_read_per_mtok": 0.50,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-sonnet-4-6": {
|
||||
"input_per_mtok": 3.00,
|
||||
"output_per_mtok": 15.00,
|
||||
"cache_read_per_mtok": 0.30,
|
||||
"cache_write_per_mtok": 3.75,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-sonnet-4-5": {
|
||||
"input_per_mtok": 3.00,
|
||||
"output_per_mtok": 15.00,
|
||||
"cache_read_per_mtok": 0.30,
|
||||
"cache_write_per_mtok": 3.75,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-haiku-4-5": {
|
||||
"input_per_mtok": 1.00,
|
||||
"output_per_mtok": 5.00,
|
||||
"cache_read_per_mtok": 0.10,
|
||||
"cache_write_per_mtok": 1.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-3-7-sonnet": {
|
||||
"input_per_mtok": 3.00,
|
||||
"output_per_mtok": 15.00,
|
||||
"cache_read_per_mtok": 0.30,
|
||||
"cache_write_per_mtok": 3.75,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-3-5-haiku": {
|
||||
"input_per_mtok": 0.80,
|
||||
"output_per_mtok": 4.00,
|
||||
"cache_read_per_mtok": 0.08,
|
||||
"cache_write_per_mtok": 1.00,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"claude-fable-5": {
|
||||
"input_per_mtok": 10.00,
|
||||
"output_per_mtok": 50.00,
|
||||
"cache_read_per_mtok": 1.00,
|
||||
"cache_write_per_mtok": 12.50,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing",
|
||||
"as_of": "2026-06"
|
||||
}
|
||||
}
|
||||
@@ -1,155 +0,0 @@
|
||||
{
|
||||
"qwen3-coder": {
|
||||
"input_per_mtok": 1.0,
|
||||
"output_per_mtok": 5.0,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Billing id maps to qwen3-coder-plus. Tiered by input length; headline is the 0-32K tier. Higher tiers: 32K-128K in 1.8/out 9, 128K-256K in 3/out 15, 256K-1M in 6/out 60. Cross-checked vs LiteLLM tiered_pricing."
|
||||
},
|
||||
"qwen-max": {
|
||||
"input_per_mtok": 1.6,
|
||||
"output_per_mtok": 6.4,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Alibaba DashScope international (USD-denominated). Cross-checked vs LiteLLM dashscope/qwen-max."
|
||||
},
|
||||
"qwen-plus": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 1.2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Alibaba DashScope international (USD). Cross-checked vs LiteLLM dashscope/qwen-plus."
|
||||
},
|
||||
"qwen-turbo": {
|
||||
"input_per_mtok": 0.05,
|
||||
"output_per_mtok": 0.2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Alibaba DashScope international (USD). Cross-checked vs LiteLLM dashscope/qwen-turbo."
|
||||
},
|
||||
"qwen3-max": {
|
||||
"input_per_mtok": 1.2,
|
||||
"output_per_mtok": 6.0,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Tiered by input length; headline is the 0-32K tier. Higher tiers: 32K-128K in 2.4/out 12, 128K-252K in 3/out 15. Cross-checked vs LiteLLM dashscope/qwen3-max tiered_pricing."
|
||||
},
|
||||
"kimi-k2": {
|
||||
"input_per_mtok": 0.6,
|
||||
"output_per_mtok": 2.5,
|
||||
"cache_read_per_mtok": 0.15,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing/chat",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Moonshot international platform (USD). Resolves to kimi-k2-0905-preview. Cache-hit input 0.15. Cross-checked vs LiteLLM moonshot/kimi-k2-0905-preview."
|
||||
},
|
||||
"kimi-k2-turbo": {
|
||||
"input_per_mtok": 1.15,
|
||||
"output_per_mtok": 8.0,
|
||||
"cache_read_per_mtok": 0.15,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing/chat",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Moonshot international (USD). Resolves to kimi-k2-turbo-preview. Cache-hit input 0.15. Cross-checked vs LiteLLM moonshot/kimi-k2-turbo-preview."
|
||||
},
|
||||
"moonshot-v1-8k": {
|
||||
"input_per_mtok": 0.2,
|
||||
"output_per_mtok": 2.0,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Moonshot international (USD). Cross-checked vs LiteLLM moonshot/moonshot-v1-8k."
|
||||
},
|
||||
"moonshot-v1-32k": {
|
||||
"input_per_mtok": 1.0,
|
||||
"output_per_mtok": 3.0,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Moonshot international (USD). Cross-checked vs LiteLLM moonshot/moonshot-v1-32k."
|
||||
},
|
||||
"moonshot-v1-128k": {
|
||||
"input_per_mtok": 2.0,
|
||||
"output_per_mtok": 5.0,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Moonshot international (USD). Cross-checked vs LiteLLM moonshot/moonshot-v1-128k."
|
||||
},
|
||||
"deepseek-v3": {
|
||||
"input_per_mtok": 0.27,
|
||||
"output_per_mtok": 1.1,
|
||||
"cache_read_per_mtok": 0.07,
|
||||
"cache_write_per_mtok": 0.0,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Legacy DeepSeek-V3 era pricing. No longer shown on the official current pricing page (which now lists deepseek-v4-flash/pro). Value from LiteLLM deepseek/deepseek-v3, last sourced from the official page. cache-hit input 0.07."
|
||||
},
|
||||
"deepseek-r1": {
|
||||
"input_per_mtok": 0.55,
|
||||
"output_per_mtok": 2.19,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Legacy DeepSeek-R1 pricing. Not on the official current pricing page (now deepseek-v4-flash/pro). Value from LiteLLM deepseek/deepseek-r1."
|
||||
},
|
||||
"deepseek-chat": {
|
||||
"input_per_mtok": 0.14,
|
||||
"output_per_mtok": 0.28,
|
||||
"cache_read_per_mtok": 0.0028,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Official page (fetched 2026-06): deepseek-chat is deprecated 2026/07/24 and now maps to deepseek-v4-flash non-thinking mode. Live billing = v4-flash: input(cache-miss) 0.14, output 0.28, input(cache-hit) 0.0028."
|
||||
},
|
||||
"deepseek-reasoner": {
|
||||
"input_per_mtok": 0.14,
|
||||
"output_per_mtok": 0.28,
|
||||
"cache_read_per_mtok": 0.0028,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Official page (fetched 2026-06): deepseek-reasoner is deprecated 2026/07/24 and now maps to deepseek-v4-flash thinking mode. Live billing = v4-flash: input(cache-miss) 0.14, output 0.28, input(cache-hit) 0.0028."
|
||||
},
|
||||
"glm-4.6": {
|
||||
"input_per_mtok": 0.6,
|
||||
"output_per_mtok": 2.2,
|
||||
"cache_read_per_mtok": 0.11,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.z.ai/guides/overview/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Z.AI international (USD). Cached-input 0.11. Cache-input storage currently limited-time free. Cross-checked vs LiteLLM zai/glm-4.6."
|
||||
},
|
||||
"glm-4-plus": {
|
||||
"input_per_mtok": null,
|
||||
"output_per_mtok": null,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": null,
|
||||
"as_of": "2026-06",
|
||||
"_note": "No authoritative current price found this session. glm-4-plus is a legacy Zhipu/BigModel-native model not listed on the Z.AI international catalog; open.bigmodel.cn pricing is SPA-rendered (no static price) and BigModel English docs returned HTTP 552/404. Not in LiteLLM. Left null per anti-hallucination rule."
|
||||
},
|
||||
"glm-4-air": {
|
||||
"input_per_mtok": 0.2,
|
||||
"output_per_mtok": 1.1,
|
||||
"cache_read_per_mtok": 0.03,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.z.ai/guides/overview/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Mission billing id glm-4-air; current Z.AI international catalog lists this as GLM-4.5-Air (USD): input 0.2, cached-input 0.03, output 1.1. Cross-checked vs LiteLLM zai/glm-4.5-air (in 0.2/out 1.1)."
|
||||
}
|
||||
}
|
||||
@@ -1,74 +0,0 @@
|
||||
{
|
||||
"gemini-2.5-pro": {
|
||||
"input_per_mtok": 1.25,
|
||||
"output_per_mtok": 10.0,
|
||||
"cache_read_per_mtok": 0.125,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Tiered: prompts >200k tokens cost input $2.50, output $15.00, cache_read $0.25 per Mtok. Cache storage $4.50/Mtok/hr."
|
||||
},
|
||||
"gemini-2.5-flash": {
|
||||
"input_per_mtok": 0.3,
|
||||
"output_per_mtok": 2.5,
|
||||
"cache_read_per_mtok": 0.03,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Text/image/video input; audio input $1.00. Output includes thinking tokens. Cache storage $1.00/Mtok/hr. No >200k context tier."
|
||||
},
|
||||
"gemini-2.5-flash-lite": {
|
||||
"input_per_mtok": 0.1,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.01,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Text/image/video input; audio input $0.30. Cache storage $1.00/Mtok/hr. No >200k context tier."
|
||||
},
|
||||
"gemini-2.0-flash": {
|
||||
"input_per_mtok": 0.1,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.025,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Text/image/video input; audio input $0.70, audio cache_read $0.175. Cache storage $1.00/Mtok/hr. No >200k context tier."
|
||||
},
|
||||
"gemini-2.0-flash-lite": {
|
||||
"input_per_mtok": 0.075,
|
||||
"output_per_mtok": 0.3,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "DEPRECATED — shut down June 1, 2026. Context caching not available for this model. Prices retained for historical billing."
|
||||
},
|
||||
"gemini-2.0-pro": {
|
||||
"input_per_mtok": null,
|
||||
"output_per_mtok": null,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "No GA billing id. Only shipped as gemini-2.0-pro-exp (experimental, free of charge); never had paid pricing. Not listed on official pricing page or LiteLLM catalog."
|
||||
},
|
||||
"gemini-3-pro-preview": {
|
||||
"input_per_mtok": 2.0,
|
||||
"output_per_mtok": 12.0,
|
||||
"cache_read_per_mtok": 0.2,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Current 3.x flagship (page labels it Gemini 3.1 Pro Preview). Tiered: prompts >200k cost input $4.00, output $18.00, cache_read $0.40 per Mtok. Cache storage $4.50/Mtok/hr."
|
||||
},
|
||||
"gemini-3-flash-preview": {
|
||||
"input_per_mtok": 0.5,
|
||||
"output_per_mtok": 3.0,
|
||||
"cache_read_per_mtok": 0.05,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Text/image/video input; audio input $1.00, audio cache_read $0.10. Cache storage $1.00/Mtok/hr. No >200k context tier on standard."
|
||||
}
|
||||
}
|
||||
@@ -1,106 +0,0 @@
|
||||
{
|
||||
"gpt-5": {
|
||||
"input_per_mtok": 1.25,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": 0.125,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-5-mini": {
|
||||
"input_per_mtok": 0.25,
|
||||
"output_per_mtok": 2,
|
||||
"cache_read_per_mtok": 0.025,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-5-nano": {
|
||||
"input_per_mtok": 0.05,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.005,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-5-codex": {
|
||||
"input_per_mtok": 1.25,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": 0.125,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-4.1": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 8,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-4.1-mini": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 1.6,
|
||||
"cache_read_per_mtok": 0.1,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-4.1-nano": {
|
||||
"input_per_mtok": 0.1,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.025,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-4o": {
|
||||
"input_per_mtok": 2.5,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": 1.25,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"gpt-4o-mini": {
|
||||
"input_per_mtok": 0.15,
|
||||
"output_per_mtok": 0.6,
|
||||
"cache_read_per_mtok": 0.075,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"o3": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 8,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"o4-mini": {
|
||||
"input_per_mtok": 1.1,
|
||||
"output_per_mtok": 4.4,
|
||||
"cache_read_per_mtok": 0.275,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"o3-mini": {
|
||||
"input_per_mtok": 1.1,
|
||||
"output_per_mtok": 4.4,
|
||||
"cache_read_per_mtok": 0.55,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
},
|
||||
"codex-mini-latest": {
|
||||
"input_per_mtok": 1.5,
|
||||
"output_per_mtok": 6,
|
||||
"cache_read_per_mtok": 0.375,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
|
||||
"as_of": "2026-06"
|
||||
}
|
||||
}
|
||||
@@ -1,137 +0,0 @@
|
||||
{
|
||||
"grok-4": {
|
||||
"input_per_mtok": 3,
|
||||
"output_per_mtok": 15,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Legacy id; xAI public pricing page now lists newer Grok versions (4.3/4.20/build-0.1). Rate verified via LiteLLM xai/ catalog (BerriAI/litellm) which sources directly from xAI."
|
||||
},
|
||||
"grok-3": {
|
||||
"input_per_mtok": 3,
|
||||
"output_per_mtok": 15,
|
||||
"cache_read_per_mtok": 0.75,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Legacy id; rate from LiteLLM xai/grok-3 catalog (sourced from xAI)."
|
||||
},
|
||||
"grok-code-fast-1": {
|
||||
"input_per_mtok": 0.2,
|
||||
"output_per_mtok": 1.5,
|
||||
"cache_read_per_mtok": 0.02,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Legacy id; rate from LiteLLM xai/grok-code-fast-1 catalog (sourced from xAI)."
|
||||
},
|
||||
"grok-2": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Legacy id (grok-2-1212); rate from LiteLLM xai/grok-2 catalog (sourced from xAI)."
|
||||
},
|
||||
"mistral-large-latest": {
|
||||
"input_per_mtok": 0.5,
|
||||
"output_per_mtok": 1.5,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "mistral-large-latest = Mistral Large 3."
|
||||
},
|
||||
"codestral-latest": {
|
||||
"input_per_mtok": 0.3,
|
||||
"output_per_mtok": 0.9,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Codestral 2; price cross-confirmed via Google Vertex partner pricing ($0.30/$0.90)."
|
||||
},
|
||||
"devstral": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Maps to devstral-medium-latest (Mistral hosted API). devstral-small is $0.10/$0.30."
|
||||
},
|
||||
"mistral-medium": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Mistral Medium 3; cross-confirmed via Google Vertex partner pricing ($0.40/$2.00)."
|
||||
},
|
||||
"llama-4-maverick": {
|
||||
"input_per_mtok": 0.27,
|
||||
"output_per_mtok": 0.85,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.together.ai/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Host-dependent. Priced via Together AI serverless. Other hosts differ (e.g. AWS Bedrock, Vertex $0.35/$1.15)."
|
||||
},
|
||||
"llama-4-scout": {
|
||||
"input_per_mtok": 0.08,
|
||||
"output_per_mtok": 0.3,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.together.ai/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Host-dependent. Priced via Together AI serverless. Vertex lists $0.25/$0.70."
|
||||
},
|
||||
"llama-3.3-70b": {
|
||||
"input_per_mtok": 0.88,
|
||||
"output_per_mtok": 0.88,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.together.ai/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "Host-dependent. Together Llama-3.3-70B-Instruct-Turbo ($0.88/$0.88, from LiteLLM together_ai source). Cheaper on DeepInfra ($0.13/$0.39); Vertex $0.72/$0.72."
|
||||
},
|
||||
"command-a": {
|
||||
"input_per_mtok": 2.5,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://cohere.com/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "command-a-03-2025."
|
||||
},
|
||||
"command-r-plus": {
|
||||
"input_per_mtok": 2.5,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://cohere.com/pricing",
|
||||
"as_of": "2026-06",
|
||||
"_note": "command-r-plus-08-2024 (legacy)."
|
||||
},
|
||||
"amazon-nova-pro": {
|
||||
"input_per_mtok": 0.8,
|
||||
"output_per_mtok": 3.2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://aws.amazon.com/bedrock/pricing/",
|
||||
"as_of": "2026-06",
|
||||
"_note": "AWS Bedrock on-demand ($0.0008/$0.0032 per 1K tokens)."
|
||||
},
|
||||
"amazon-nova-lite": {
|
||||
"input_per_mtok": 0.06,
|
||||
"output_per_mtok": 0.24,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://aws.amazon.com/bedrock/pricing/",
|
||||
"as_of": "2026-06",
|
||||
"_note": "AWS Bedrock on-demand ($0.00006/$0.00024 per 1K tokens)."
|
||||
}
|
||||
}
|
||||
+36
-5
@@ -8,7 +8,7 @@
|
||||
import {
|
||||
lookupPrice as catalogLookupPrice,
|
||||
computeCostUsd as catalogComputeCostUsd,
|
||||
} from "../pricing/catalog.js";
|
||||
} from "./pricing.js";
|
||||
|
||||
// ── Public interfaces ──────────────────────────────────────────────────────
|
||||
|
||||
@@ -30,6 +30,19 @@ export interface SessionEvent {
|
||||
* `Fetched and indexed N sections (XKB)` preamble.
|
||||
*/
|
||||
bytes_avoided?: number;
|
||||
/**
|
||||
* Optional structured cost/usage fields (Wave 2b). Emitted by
|
||||
* extractAgentUsage alongside the colon-string `data` so the forward
|
||||
* envelope can spread them to the platform as typed columns instead of an
|
||||
* opaque blob. Present only when the source signal is present; cost_usd is
|
||||
* omitted on a price miss or a zero-token turn.
|
||||
*/
|
||||
model_id?: string;
|
||||
input_tokens?: number;
|
||||
output_tokens?: number;
|
||||
cache_read_tokens?: number;
|
||||
cache_creation_tokens?: number;
|
||||
cost_usd?: number;
|
||||
}
|
||||
|
||||
export interface ToolCall {
|
||||
@@ -1487,21 +1500,39 @@ function extractAgentUsage(input: HookInput): SessionEvent[] {
|
||||
const cacheRead = typeof usage.cache_read_input_tokens === "number"
|
||||
? usage.cache_read_input_tokens
|
||||
: 0;
|
||||
const modelId = resolveModelId(input, out);
|
||||
const anyTokens = inputTokens > 0 || outputTokens > 0 || cacheCreate > 0 || cacheRead > 0;
|
||||
let cost: number | null = null;
|
||||
if (anyTokens) {
|
||||
const modelId = resolveModelId(input, out);
|
||||
// null ⇒ unmatched model id (catalog warned once) — skip the cost token
|
||||
// rather than blend a wrong Claude rate (the old non-Claude bug).
|
||||
const cost = computeTurnCostUsd(modelId, inputTokens, outputTokens, cacheCreate, cacheRead);
|
||||
cost = computeTurnCostUsd(modelId, inputTokens, outputTokens, cacheCreate, cacheRead);
|
||||
if (cost !== null) parts.push(`cost_usd:${formatCostUsd(cost)}`);
|
||||
}
|
||||
|
||||
return [{
|
||||
// Wave 2b — emit structured top-level fields alongside the colon-string so
|
||||
// the forward envelope (which spreads `...event`) hands the platform typed
|
||||
// columns. Each field is set only when its source signal is present, so the
|
||||
// forward payload stays minimal; cost_usd is omitted on a price miss or a
|
||||
// zero-token turn. The colon-string `data` stays for human/debug + back-compat.
|
||||
const event: SessionEvent = {
|
||||
type: "agent_usage",
|
||||
category: "cost",
|
||||
data: safeString(parts.join(" ")),
|
||||
priority: 2,
|
||||
}];
|
||||
};
|
||||
if (modelId.length > 0) event.model_id = modelId;
|
||||
if (typeof usage.input_tokens === "number") event.input_tokens = usage.input_tokens;
|
||||
if (typeof usage.output_tokens === "number") event.output_tokens = usage.output_tokens;
|
||||
if (typeof usage.cache_read_input_tokens === "number") {
|
||||
event.cache_read_tokens = usage.cache_read_input_tokens;
|
||||
}
|
||||
if (typeof usage.cache_creation_input_tokens === "number") {
|
||||
event.cache_creation_tokens = usage.cache_creation_input_tokens;
|
||||
}
|
||||
if (cost !== null) event.cost_usd = cost;
|
||||
|
||||
return [event];
|
||||
}
|
||||
|
||||
// ── User-message extractors ────────────────────────────────────────────────
|
||||
|
||||
@@ -0,0 +1,429 @@
|
||||
{
|
||||
"claude-opus-4-8": {
|
||||
"input_per_mtok": 5,
|
||||
"output_per_mtok": 25,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-opus-4-7": {
|
||||
"input_per_mtok": 5,
|
||||
"output_per_mtok": 25,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-opus-4-6": {
|
||||
"input_per_mtok": 5,
|
||||
"output_per_mtok": 25,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-opus-4-5": {
|
||||
"input_per_mtok": 5,
|
||||
"output_per_mtok": 25,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": 6.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-sonnet-4-6": {
|
||||
"input_per_mtok": 3,
|
||||
"output_per_mtok": 15,
|
||||
"cache_read_per_mtok": 0.3,
|
||||
"cache_write_per_mtok": 3.75,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-sonnet-4-5": {
|
||||
"input_per_mtok": 3,
|
||||
"output_per_mtok": 15,
|
||||
"cache_read_per_mtok": 0.3,
|
||||
"cache_write_per_mtok": 3.75,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-haiku-4-5": {
|
||||
"input_per_mtok": 1,
|
||||
"output_per_mtok": 5,
|
||||
"cache_read_per_mtok": 0.1,
|
||||
"cache_write_per_mtok": 1.25,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-3-7-sonnet": {
|
||||
"input_per_mtok": 3,
|
||||
"output_per_mtok": 15,
|
||||
"cache_read_per_mtok": 0.3,
|
||||
"cache_write_per_mtok": 3.75,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"claude-3-5-haiku": {
|
||||
"input_per_mtok": 0.8,
|
||||
"output_per_mtok": 4,
|
||||
"cache_read_per_mtok": 0.08,
|
||||
"cache_write_per_mtok": 1,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"claude-fable-5": {
|
||||
"input_per_mtok": 10,
|
||||
"output_per_mtok": 50,
|
||||
"cache_read_per_mtok": 1,
|
||||
"cache_write_per_mtok": 12.5,
|
||||
"source": "https://platform.claude.com/docs/en/about-claude/pricing"
|
||||
},
|
||||
"gpt-5": {
|
||||
"input_per_mtok": 1.25,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": 0.125,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-5-mini": {
|
||||
"input_per_mtok": 0.25,
|
||||
"output_per_mtok": 2,
|
||||
"cache_read_per_mtok": 0.025,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-5-nano": {
|
||||
"input_per_mtok": 0.05,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.005,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-5-codex": {
|
||||
"input_per_mtok": 1.25,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": 0.125,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-4.1": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 8,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-4.1-mini": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 1.6,
|
||||
"cache_read_per_mtok": 0.1,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-4.1-nano": {
|
||||
"input_per_mtok": 0.1,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.025,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-4o": {
|
||||
"input_per_mtok": 2.5,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": 1.25,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gpt-4o-mini": {
|
||||
"input_per_mtok": 0.15,
|
||||
"output_per_mtok": 0.6,
|
||||
"cache_read_per_mtok": 0.075,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"o3": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 8,
|
||||
"cache_read_per_mtok": 0.5,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"o4-mini": {
|
||||
"input_per_mtok": 1.1,
|
||||
"output_per_mtok": 4.4,
|
||||
"cache_read_per_mtok": 0.275,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"o3-mini": {
|
||||
"input_per_mtok": 1.1,
|
||||
"output_per_mtok": 4.4,
|
||||
"cache_read_per_mtok": 0.55,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"codex-mini-latest": {
|
||||
"input_per_mtok": 1.5,
|
||||
"output_per_mtok": 6,
|
||||
"cache_read_per_mtok": 0.375,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
|
||||
},
|
||||
"gemini-2.5-pro": {
|
||||
"input_per_mtok": 1.25,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": 0.125,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing"
|
||||
},
|
||||
"gemini-2.5-flash": {
|
||||
"input_per_mtok": 0.3,
|
||||
"output_per_mtok": 2.5,
|
||||
"cache_read_per_mtok": 0.03,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing"
|
||||
},
|
||||
"gemini-2.5-flash-lite": {
|
||||
"input_per_mtok": 0.1,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.01,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing"
|
||||
},
|
||||
"gemini-2.0-flash": {
|
||||
"input_per_mtok": 0.1,
|
||||
"output_per_mtok": 0.4,
|
||||
"cache_read_per_mtok": 0.025,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing"
|
||||
},
|
||||
"gemini-2.0-flash-lite": {
|
||||
"input_per_mtok": 0.075,
|
||||
"output_per_mtok": 0.3,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing"
|
||||
},
|
||||
"gemini-3-pro-preview": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 12,
|
||||
"cache_read_per_mtok": 0.2,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing"
|
||||
},
|
||||
"gemini-3-flash-preview": {
|
||||
"input_per_mtok": 0.5,
|
||||
"output_per_mtok": 3,
|
||||
"cache_read_per_mtok": 0.05,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://ai.google.dev/gemini-api/docs/pricing"
|
||||
},
|
||||
"qwen3-coder": {
|
||||
"input_per_mtok": 1,
|
||||
"output_per_mtok": 5,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models"
|
||||
},
|
||||
"qwen-max": {
|
||||
"input_per_mtok": 1.6,
|
||||
"output_per_mtok": 6.4,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models"
|
||||
},
|
||||
"qwen-plus": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 1.2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models"
|
||||
},
|
||||
"qwen-turbo": {
|
||||
"input_per_mtok": 0.05,
|
||||
"output_per_mtok": 0.2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models"
|
||||
},
|
||||
"qwen3-max": {
|
||||
"input_per_mtok": 1.2,
|
||||
"output_per_mtok": 6,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.alibabacloud.com/help/en/model-studio/models"
|
||||
},
|
||||
"kimi-k2": {
|
||||
"input_per_mtok": 0.6,
|
||||
"output_per_mtok": 2.5,
|
||||
"cache_read_per_mtok": 0.15,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing/chat"
|
||||
},
|
||||
"kimi-k2-turbo": {
|
||||
"input_per_mtok": 1.15,
|
||||
"output_per_mtok": 8,
|
||||
"cache_read_per_mtok": 0.15,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing/chat"
|
||||
},
|
||||
"moonshot-v1-8k": {
|
||||
"input_per_mtok": 0.2,
|
||||
"output_per_mtok": 2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing"
|
||||
},
|
||||
"moonshot-v1-32k": {
|
||||
"input_per_mtok": 1,
|
||||
"output_per_mtok": 3,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing"
|
||||
},
|
||||
"moonshot-v1-128k": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 5,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://platform.moonshot.ai/docs/pricing"
|
||||
},
|
||||
"deepseek-v3": {
|
||||
"input_per_mtok": 0.27,
|
||||
"output_per_mtok": 1.1,
|
||||
"cache_read_per_mtok": 0.07,
|
||||
"cache_write_per_mtok": 0,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing"
|
||||
},
|
||||
"deepseek-r1": {
|
||||
"input_per_mtok": 0.55,
|
||||
"output_per_mtok": 2.19,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing"
|
||||
},
|
||||
"deepseek-chat": {
|
||||
"input_per_mtok": 0.14,
|
||||
"output_per_mtok": 0.28,
|
||||
"cache_read_per_mtok": 0.0028,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing"
|
||||
},
|
||||
"deepseek-reasoner": {
|
||||
"input_per_mtok": 0.14,
|
||||
"output_per_mtok": 0.28,
|
||||
"cache_read_per_mtok": 0.0028,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://api-docs.deepseek.com/quick_start/pricing"
|
||||
},
|
||||
"glm-4.6": {
|
||||
"input_per_mtok": 0.6,
|
||||
"output_per_mtok": 2.2,
|
||||
"cache_read_per_mtok": 0.11,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.z.ai/guides/overview/pricing"
|
||||
},
|
||||
"glm-4-air": {
|
||||
"input_per_mtok": 0.2,
|
||||
"output_per_mtok": 1.1,
|
||||
"cache_read_per_mtok": 0.03,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.z.ai/guides/overview/pricing"
|
||||
},
|
||||
"grok-4": {
|
||||
"input_per_mtok": 3,
|
||||
"output_per_mtok": 15,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing"
|
||||
},
|
||||
"grok-3": {
|
||||
"input_per_mtok": 3,
|
||||
"output_per_mtok": 15,
|
||||
"cache_read_per_mtok": 0.75,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing"
|
||||
},
|
||||
"grok-code-fast-1": {
|
||||
"input_per_mtok": 0.2,
|
||||
"output_per_mtok": 1.5,
|
||||
"cache_read_per_mtok": 0.02,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing"
|
||||
},
|
||||
"grok-2": {
|
||||
"input_per_mtok": 2,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://docs.x.ai/docs/pricing"
|
||||
},
|
||||
"mistral-large-latest": {
|
||||
"input_per_mtok": 0.5,
|
||||
"output_per_mtok": 1.5,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing"
|
||||
},
|
||||
"codestral-latest": {
|
||||
"input_per_mtok": 0.3,
|
||||
"output_per_mtok": 0.9,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing"
|
||||
},
|
||||
"devstral": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing"
|
||||
},
|
||||
"mistral-medium": {
|
||||
"input_per_mtok": 0.4,
|
||||
"output_per_mtok": 2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://mistral.ai/pricing"
|
||||
},
|
||||
"llama-4-maverick": {
|
||||
"input_per_mtok": 0.27,
|
||||
"output_per_mtok": 0.85,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.together.ai/pricing"
|
||||
},
|
||||
"llama-4-scout": {
|
||||
"input_per_mtok": 0.08,
|
||||
"output_per_mtok": 0.3,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.together.ai/pricing"
|
||||
},
|
||||
"llama-3.3-70b": {
|
||||
"input_per_mtok": 0.88,
|
||||
"output_per_mtok": 0.88,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://www.together.ai/pricing"
|
||||
},
|
||||
"command-a": {
|
||||
"input_per_mtok": 2.5,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://cohere.com/pricing"
|
||||
},
|
||||
"command-r-plus": {
|
||||
"input_per_mtok": 2.5,
|
||||
"output_per_mtok": 10,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://cohere.com/pricing"
|
||||
},
|
||||
"amazon-nova-pro": {
|
||||
"input_per_mtok": 0.8,
|
||||
"output_per_mtok": 3.2,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://aws.amazon.com/bedrock/pricing/"
|
||||
},
|
||||
"amazon-nova-lite": {
|
||||
"input_per_mtok": 0.06,
|
||||
"output_per_mtok": 0.24,
|
||||
"cache_read_per_mtok": null,
|
||||
"cache_write_per_mtok": null,
|
||||
"source": "https://aws.amazon.com/bedrock/pricing/"
|
||||
}
|
||||
}
|
||||
@@ -1,9 +1,9 @@
|
||||
/**
|
||||
* Pricing catalog — single source of truth for per-model USD cost.
|
||||
*
|
||||
* Deep module, tiny interface. At load it merges the 5 curated vendor JSONs
|
||||
* (src/pricing/sources/{anthropic,openai,google,chinese,others}.json) into one
|
||||
* Map<modelId, Price> in per-Mtok units, then exposes three pure-ish functions:
|
||||
* Deep module, tiny interface. At load it reads one curated multi-vendor JSON
|
||||
* (src/session/model-prices.json) into a Map<modelId, Price> in per-Mtok units,
|
||||
* then exposes three pure-ish functions:
|
||||
*
|
||||
* lookupPrice(modelId) → Price | null
|
||||
* computeCostUsd(modelId, tokens) → number | null
|
||||
@@ -17,18 +17,14 @@
|
||||
* an unknown id resolves to `null` (no price) instead of a wrong Claude rate.
|
||||
*
|
||||
* The large litellm catalog (~1.5MB, ~2900 models) is NOT bundled — it lives at
|
||||
* tools/pricing/litellm-catalog.json as the dev-only refresh base for these
|
||||
* curated JSONs. See tools/pricing/litellm-NOTES.md for the refresh recipe.
|
||||
* tools/pricing/litellm-catalog.json as the dev-only refresh base for this
|
||||
* curated JSON.
|
||||
*
|
||||
* The 5 curated JSONs are small (~25KB total) and esbuild inlines them into the
|
||||
* The curated JSON is small (~13KB, 61 models) and esbuild inlines it into the
|
||||
* hook/server bundles at build time (no runtime fs read, no external file).
|
||||
*/
|
||||
|
||||
import anthropic from "./sources/anthropic.json" with { type: "json" };
|
||||
import openai from "./sources/openai.json" with { type: "json" };
|
||||
import google from "./sources/google.json" with { type: "json" };
|
||||
import chinese from "./sources/chinese.json" with { type: "json" };
|
||||
import others from "./sources/others.json" with { type: "json" };
|
||||
import catalog from "./model-prices.json" with { type: "json" };
|
||||
|
||||
/** Per-Mtok price for one model. Any of the four rates may be null ("unknown"). */
|
||||
export interface Price {
|
||||
@@ -46,7 +42,7 @@ export interface TokenCounts {
|
||||
cache_creation_tokens?: number;
|
||||
}
|
||||
|
||||
/** Raw shape of a curated source row (carries provenance fields we drop). */
|
||||
/** Raw shape of a curated source row (carries a provenance `source` we drop). */
|
||||
interface RawRow {
|
||||
input_per_mtok: number | null;
|
||||
output_per_mtok: number | null;
|
||||
@@ -56,36 +52,28 @@ interface RawRow {
|
||||
}
|
||||
|
||||
/**
|
||||
* Merge the 5 vendor JSONs into one Map. A row with a null *input* price is
|
||||
* Read the curated JSON into one Map. A row with a null *input* price is
|
||||
* unusable for cost (the primary bucket has no rate) and is dropped at load so
|
||||
* lookupPrice returns null for it — matching "null-priced entries → no price".
|
||||
* Curated ids are globally unique across the five files (verified), so merge
|
||||
* order is irrelevant; later files would otherwise win on collision.
|
||||
* (The two null-input ids are already pruned from the JSON itself; this guard
|
||||
* keeps the loader robust if one is ever re-added.)
|
||||
*/
|
||||
function buildCatalog(): Map<string, Price> {
|
||||
const map = new Map<string, Price>();
|
||||
const sources: Record<string, RawRow>[] = [
|
||||
anthropic as Record<string, RawRow>,
|
||||
openai as Record<string, RawRow>,
|
||||
google as Record<string, RawRow>,
|
||||
chinese as Record<string, RawRow>,
|
||||
others as Record<string, RawRow>,
|
||||
];
|
||||
for (const src of sources) {
|
||||
for (const id of Object.keys(src)) {
|
||||
const row = src[id];
|
||||
if (row == null || typeof row !== "object") continue;
|
||||
// No input rate ⇒ no usable price for this model.
|
||||
if (typeof row.input_per_mtok !== "number") continue;
|
||||
map.set(id, {
|
||||
input_per_mtok: row.input_per_mtok,
|
||||
output_per_mtok: typeof row.output_per_mtok === "number" ? row.output_per_mtok : null,
|
||||
cache_read_per_mtok:
|
||||
typeof row.cache_read_per_mtok === "number" ? row.cache_read_per_mtok : null,
|
||||
cache_write_per_mtok:
|
||||
typeof row.cache_write_per_mtok === "number" ? row.cache_write_per_mtok : null,
|
||||
});
|
||||
}
|
||||
const src = catalog as Record<string, RawRow>;
|
||||
for (const id of Object.keys(src)) {
|
||||
const row = src[id];
|
||||
if (row == null || typeof row !== "object") continue;
|
||||
// No input rate ⇒ no usable price for this model.
|
||||
if (typeof row.input_per_mtok !== "number") continue;
|
||||
map.set(id, {
|
||||
input_per_mtok: row.input_per_mtok,
|
||||
output_per_mtok: typeof row.output_per_mtok === "number" ? row.output_per_mtok : null,
|
||||
cache_read_per_mtok:
|
||||
typeof row.cache_read_per_mtok === "number" ? row.cache_read_per_mtok : null,
|
||||
cache_write_per_mtok:
|
||||
typeof row.cache_write_per_mtok === "number" ? row.cache_write_per_mtok : null,
|
||||
});
|
||||
}
|
||||
return map;
|
||||
}
|
||||
@@ -292,3 +292,90 @@ describe("extractAgentUsage — Issue #4 AgentOutput.usage capture", () => {
|
||||
}
|
||||
});
|
||||
});
|
||||
|
||||
/**
|
||||
* Wave 2b — structured cost event.
|
||||
*
|
||||
* The colon-string `data` is opaque to the platform (it cannot column-ize a
|
||||
* "tokens_in:123 cost_usd:0.02" blob). extractAgentUsage now also emits the
|
||||
* cost/token signals as top-level SessionEvent fields, which the forward
|
||||
* envelope spreads straight to the platform as typed columns:
|
||||
*
|
||||
* model_id, input_tokens, output_tokens,
|
||||
* cache_read_tokens, cache_creation_tokens, cost_usd
|
||||
*
|
||||
* The colon-string `data` stays for human/debug + back-compat.
|
||||
*/
|
||||
describe("extractAgentUsage — Wave 2b structured cost fields", () => {
|
||||
function usageEvent(toolInput: Record<string, unknown>, usage: Record<string, unknown>, extra: Record<string, unknown> = {}) {
|
||||
return extractEvents({
|
||||
tool_name: "Task",
|
||||
tool_input: toolInput,
|
||||
tool_response: JSON.stringify({ ...extra, usage }),
|
||||
}).filter((e) => e.type === "agent_usage")[0];
|
||||
}
|
||||
|
||||
// (a) the 6 structured fields ride the event with correct values
|
||||
test("(a) Task usage yields event carrying the 6 structured fields with correct values", () => {
|
||||
const ev = usageEvent(
|
||||
{ model: "claude-sonnet-4-6" },
|
||||
{
|
||||
input_tokens: 1000,
|
||||
output_tokens: 500,
|
||||
cache_creation_input_tokens: 1000,
|
||||
cache_read_input_tokens: 1500,
|
||||
},
|
||||
);
|
||||
expect(ev.model_id).toBe("claude-sonnet-4-6");
|
||||
expect(ev.input_tokens).toBe(1000);
|
||||
expect(ev.output_tokens).toBe(500);
|
||||
expect(ev.cache_creation_tokens).toBe(1000);
|
||||
expect(ev.cache_read_tokens).toBe(1500);
|
||||
// 1000*3 + 500*15 + 1000*3.75 + 1500*0.30 = 14700 / 1e6 = 0.0147
|
||||
expect(ev.cost_usd).toBeCloseTo(0.0147, 8);
|
||||
});
|
||||
|
||||
// (b) cost_usd matches the catalog for a known model
|
||||
test("(b) cost_usd matches the catalog for a known model (gpt-5)", () => {
|
||||
const ev = usageEvent(
|
||||
{ model: "gpt-5" },
|
||||
{ input_tokens: 1000, output_tokens: 500 },
|
||||
);
|
||||
// gpt-5: 1000*1.25 + 500*10 = 6250 / 1e6 = 0.00625
|
||||
expect(ev.cost_usd).toBeCloseTo(0.00625, 8);
|
||||
expect(ev.model_id).toBe("gpt-5");
|
||||
});
|
||||
|
||||
// (c) unknown model → tokens present, cost_usd omitted (no Claude fallback)
|
||||
test("(c) unknown model → tokens present, cost_usd omitted/null", () => {
|
||||
const warn = vi.spyOn(console, "warn").mockImplementation(() => {});
|
||||
const ev = usageEvent(
|
||||
{ model: "claude-future-model-99" },
|
||||
{ input_tokens: 1000, output_tokens: 500 },
|
||||
);
|
||||
warn.mockRestore();
|
||||
expect(ev.input_tokens).toBe(1000);
|
||||
expect(ev.output_tokens).toBe(500);
|
||||
expect(ev.cost_usd == null).toBe(true);
|
||||
});
|
||||
|
||||
// (d) zero-token response → no cost_usd
|
||||
test("(d) zero-token response → no cost_usd", () => {
|
||||
const ev = usageEvent(
|
||||
{ model: "claude-sonnet-4-6" },
|
||||
{ input_tokens: 0, output_tokens: 0 },
|
||||
);
|
||||
expect(ev.cost_usd == null).toBe(true);
|
||||
});
|
||||
|
||||
// (e) the existing colon-string `data` still present for back-compat
|
||||
test("(e) colon-string data still present for back-compat", () => {
|
||||
const ev = usageEvent(
|
||||
{ model: "claude-sonnet-4-6" },
|
||||
{ input_tokens: 1000, output_tokens: 500 },
|
||||
);
|
||||
expect(ev.data).toMatch(/tokens_in:1000/);
|
||||
expect(ev.data).toMatch(/tokens_out:500/);
|
||||
expect(ev.data).toMatch(/cost_usd:0\.0105/);
|
||||
});
|
||||
});
|
||||
|
||||
@@ -6,7 +6,7 @@
|
||||
* a Claude rate (non-Claude models silently inherited Sonnet's default),
|
||||
* which over-/under-charged every OpenAI / Gemini / Qwen / DeepSeek turn.
|
||||
*
|
||||
* The catalog merges the 5 curated vendor JSONs (src/pricing/sources/*.json)
|
||||
* The catalog reads the curated multi-vendor JSON (src/session/model-prices.json)
|
||||
* into one per-Mtok price map and prices each model from ITS OWN row.
|
||||
* Behaviours under test:
|
||||
* (a) curated lookup hits across all five vendors
|
||||
@@ -23,9 +23,9 @@ import {
|
||||
lookupPrice,
|
||||
computeCostUsd,
|
||||
nativeOrComputed,
|
||||
} from "../../src/pricing/catalog.js";
|
||||
} from "../../src/session/pricing.js";
|
||||
|
||||
describe("pricing/catalog — lookupPrice", () => {
|
||||
describe("session/pricing — lookupPrice", () => {
|
||||
// (a) curated lookup hits
|
||||
test("(a) Anthropic curated hit returns per-Mtok price", () => {
|
||||
const p = lookupPrice("claude-opus-4-8");
|
||||
@@ -79,7 +79,7 @@ describe("pricing/catalog — lookupPrice", () => {
|
||||
});
|
||||
});
|
||||
|
||||
describe("pricing/catalog — computeCostUsd", () => {
|
||||
describe("session/pricing — computeCostUsd", () => {
|
||||
// (c) THE BUG: non-Claude model must use its own price, not Claude's.
|
||||
test("(c) gpt-5 priced from its own row, NOT Claude default", () => {
|
||||
const tokens = { input_tokens: 1000, output_tokens: 500 };
|
||||
@@ -157,7 +157,7 @@ describe("pricing/catalog — computeCostUsd", () => {
|
||||
});
|
||||
});
|
||||
|
||||
describe("pricing/catalog — nativeOrComputed", () => {
|
||||
describe("session/pricing — nativeOrComputed", () => {
|
||||
// (g) native-cost passthrough
|
||||
test("(g) provider native cost wins over computed", () => {
|
||||
const native = nativeOrComputed("gpt-5", { input_tokens: 1000 }, 0.42);
|
||||
@@ -1,80 +0,0 @@
|
||||
# LiteLLM Catalog — Adapter Notes
|
||||
|
||||
Vendored from BerriAI/litellm as context-mode's comprehensive pricing base, so
|
||||
unknown/unseen models still resolve a price instead of failing.
|
||||
|
||||
- **Source:** `https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json`
|
||||
- **Local file:** `litellm-catalog.json`
|
||||
- **Size:** ~1.53 MB (1,570,159 bytes)
|
||||
- **Top-level keys:** 2,901 (2,900 model entries + 1 `sample_spec` template — skip `sample_spec`)
|
||||
- **Entries with numeric `input_cost_per_token`:** 2,432
|
||||
- **Distinct `litellm_provider` values:** 120 (top: fireworks_ai, bedrock, openai, azure, gemini, mistral, openrouter …)
|
||||
|
||||
## Shape
|
||||
|
||||
The file is a flat JSON object. **Each key is a model id**; each value is a metadata object.
|
||||
There is no wrapper array. One special key, `sample_spec`, is a documentation template
|
||||
(not a real model) and must be excluded.
|
||||
|
||||
```jsonc
|
||||
{
|
||||
"sample_spec": { /* template — ignore */ },
|
||||
"gpt-4o": { "input_cost_per_token": 0.0000025, ... },
|
||||
"claude-sonnet-4-20250514": { ... }
|
||||
}
|
||||
```
|
||||
|
||||
## Cost fields — verified present in the fetched JSON
|
||||
|
||||
> All `*_cost_per_token` values are **PER TOKEN** (USD), not per-million.
|
||||
> To convert to per-Mtok (context-mode's internal unit): **multiply by `1e6` (×1,000,000).**
|
||||
> e.g. `gpt-4o` `input_cost_per_token = 0.0000025` → `0.0000025 × 1e6 = $2.50 / Mtok`.
|
||||
|
||||
Core fields (use these; coverage count in parentheses):
|
||||
|
||||
| Field | Meaning | Count |
|
||||
|-------|---------|-------|
|
||||
| `input_cost_per_token` | prompt cost per token | 2,433 |
|
||||
| `output_cost_per_token` | completion cost per token | 2,430 |
|
||||
| `cache_read_input_token_cost` | cached-prompt read cost per token | 645 |
|
||||
| `cache_creation_input_token_cost` | cache-write/creation cost per token | 210 |
|
||||
|
||||
Real example (`claude-sonnet-4-20250514`, provider `anthropic`):
|
||||
`input_cost_per_token: 0.000003` (= $3.00/Mtok), `output_cost_per_token: 0.000015` (= $15.00/Mtok),
|
||||
`cache_read_input_token_cost: 3e-7` (= $0.30/Mtok), `cache_creation_input_token_cost: 0.00000375` (= $3.75/Mtok).
|
||||
|
||||
### Schema surprises / gotchas
|
||||
|
||||
- **Tiered & variant cost fields exist** — e.g. `input_cost_per_token_above_200k_tokens`,
|
||||
`cache_read_input_token_cost_above_200k_tokens`, `_priority`, `_batches`, `_flex`,
|
||||
`_above_1hr`, `_above_272k_tokens`. Treat these as optional overrides; fall back to the
|
||||
base `input_cost_per_token` / `output_cost_per_token`.
|
||||
- **Non-token cost units also appear** and are NOT per-token: `input_cost_per_second`,
|
||||
`input_cost_per_character`, `input_cost_per_image`, `input_cost_per_pixel`,
|
||||
`output_cost_per_second`, `output_cost_per_reasoning_token`,
|
||||
`search_context_cost_per_query` (the last can be an object keyed by search-context size).
|
||||
Do not blindly ×1e6 these — only the `*_per_token` family is per-token.
|
||||
- **2,366 of 2,900 keys contain `/`** — namespaced ids like `bedrock/...`,
|
||||
`1024-x-1024/dall-e-2`, image/resolution-prefixed entries. Match on the full key.
|
||||
- Metadata fields used for context limits: `max_tokens`, `max_input_tokens`,
|
||||
`max_output_tokens` (and `mode` distinguishes `chat`, `embedding`, image, etc.).
|
||||
- Some entries carry `deprecation_date` (81 entries).
|
||||
|
||||
## Lookup strategy (for the adapter)
|
||||
|
||||
Given an incoming `model_id`:
|
||||
|
||||
1. **Exact match** — `catalog[model_id]`. Fastest; covers the common case.
|
||||
2. **Provider-stripped / namespaced fallback** — many ids are `provider/model`.
|
||||
Try stripping or adding a known provider prefix:
|
||||
- if `model_id` has no `/`, try `catalog[provider + "/" + model_id]`;
|
||||
- if `model_id` is `provider/model`, also try the bare `model` segment.
|
||||
3. **Provider + model match** — scan entries whose `litellm_provider` matches the resolved
|
||||
provider and whose key endsWith the model segment.
|
||||
4. **Skip `sample_spec`** in every scan.
|
||||
5. From the matched entry, read `input_cost_per_token` / `output_cost_per_token`
|
||||
(+ optional `cache_read_input_token_cost`, `cache_creation_input_token_cost`),
|
||||
then **× 1e6** to get per-Mtok rates.
|
||||
6. If nothing matches or `input_cost_per_token` is absent (some entries price only by
|
||||
second/character/image), the model has no usable per-token price — fall through to
|
||||
context-mode's own default.
|
||||
Reference in New Issue
Block a user