Skip to content

Commit 4e85eac

Browse files
rekram1-noderekram1-node
committed
docs: document provider reasoning request formats
1 parent d6e5057 commit 4e85eac

533 files changed

Lines changed: 2155 additions & 6 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

‎providers/302ai/provider.toml‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,11 @@
11
name = "302.AI"
22
env = ["302AI_API_KEY"]
33
npm = "@ai-sdk/openai-compatible"
4+
# Reasoning HTTP format (accessed 2026-06-25):
5+
# Audited POST https://api.302.ai/v1/chat/completions. The provider's API guide
6+
# documents model/messages only; no reasoning toggle, effort, or numeric budget
7+
# request field is documented. Do not infer passthrough from upstream APIs.
8+
# Sources:
9+
# https://doc.302.ai/
410
doc = "https://doc.302.ai"
511
api = "https://api.302.ai/v1"

‎providers/abacus/provider.toml‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,11 @@
11
name = "Abacus"
22
npm = "@ai-sdk/openai-compatible"
3+
# Reasoning HTTP format (accessed 2026-06-25):
4+
# Audited POST https://routellm.abacus.ai/v1/chat/completions. The provider API
5+
# reference documents no reasoning toggle, effort, or numeric budget request
6+
# field. Do not infer behavior from the routed model developer's API.
7+
# Sources:
8+
# https://abacus.ai/help/api
39
env = ["ABACUS_API_KEY"]
410
doc = "https://abacus.ai/help/api"
511
api = "https://routellm.abacus.ai/v1"

‎providers/abliteration-ai/models/abliterated-model.toml‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,10 @@
11
name = "Abliterated Model"
2+
# Reasoning HTTP format (accessed 2026-06-25):
3+
# This model thinks by default. On POST /v1/chat/completions or /v1/messages,
4+
# top-level `thinking: false` skips thinking; omission keeps it enabled.
5+
# Sources:
6+
# https://docs.abliteration.ai/models
7+
# https://docs.abliteration.ai/capabilities/thinking
28
release_date = "2026-01-06"
39
last_updated = "2026-01-06"
410
attachment = true

‎providers/abliteration-ai/provider.toml‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,12 @@
11
name = "abliteration.ai"
22
env = ["ABLIT_KEY"]
33
npm = "@ai-sdk/openai-compatible"
4+
# Reasoning HTTP format (accessed 2026-06-25):
5+
# POST /v1/chat/completions and POST /v1/messages: top-level `thinking` is true
6+
# by default; false skips thinking. POST /v1/responses has no thinking toggle.
7+
# No effort or numeric reasoning-budget request field is documented.
8+
# Sources:
9+
# https://docs.abliteration.ai/capabilities/thinking
10+
# https://docs.abliteration.ai/compatibility-matrix
411
api = "https://api.abliteration.ai/v1"
512
doc = "https://docs.abliteration.ai/models"

‎providers/aihubmix/models/claude-opus-4-6.toml‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -5,6 +5,7 @@ last_updated = "2026-03-13"
55
attachment = true
66
reasoning = true
77
reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
8+
# Native Messages prefers $.thinking.type = "adaptive" with $.output_config.effort = "low"|"medium"|"high"|"max"; enabled budget_tokens >= 1024 is deprecated and must be < $.max_tokens. https://docs.aihubmix.com/cn/api/Claude-Native (accessed 2026-06-25)
89
temperature = true
910
tool_call = true
1011
structured_output = true

‎providers/aihubmix/models/claude-opus-4-7.toml‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -5,6 +5,7 @@ last_updated = "2026-04-16"
55
attachment = true
66
reasoning = true
77
reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
8+
# Native Messages uses $.thinking.type = "adaptive" and $.output_config.effort = "low"|"medium"|"high"|"xhigh"|"max"; manual budget_tokens is rejected. https://docs.aihubmix.com/cn/blogs/Claude-Opus4.7 (accessed 2026-06-25)
89
temperature = false
910
tool_call = true
1011
structured_output = true

‎providers/aihubmix/models/claude-sonnet-4-6.toml‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -5,6 +5,7 @@ last_updated = "2026-03-13"
55
attachment = true
66
reasoning = true
77
reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high", "max"] }, { type = "budget_tokens", min = 1_024 }]
8+
# Native Messages prefers $.thinking.type = "adaptive" with $.output_config.effort = "low"|"medium"|"high"; enabled budget_tokens >= 1024 is deprecated and must be < $.max_tokens. Chat effort "max" maps to native "high". https://docs.aihubmix.com/cn/api/Claude-Native (accessed 2026-06-25)
89
temperature = true
910
tool_call = true
1011
structured_output = true

‎providers/aihubmix/models/gemini-2.5-flash.toml‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -5,6 +5,7 @@ last_updated = "2025-06-05"
55
attachment = true
66
reasoning = true
77
reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "high"] }, { type = "budget_tokens", min = 0, max = 24_576 }]
8+
# Native Gemini uses $.generationConfig.thinkingConfig.thinkingBudget: 0 disables, -1 is dynamic, and manual budgets are 1..24576. https://cloud.google.com/vertex-ai/generative-ai/docs/thinking (accessed 2026-06-25)
89
temperature = true
910
tool_call = true
1011
structured_output = true

‎providers/aihubmix/models/gemini-2.5-pro.toml‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -5,6 +5,7 @@ last_updated = "2025-06-05"
55
attachment = true
66
reasoning = true
77
reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
8+
# Native Gemini uses $.generationConfig.thinkingConfig.thinkingBudget: -1 is dynamic and manual budgets are 128..32768; 0/off is unsupported. https://cloud.google.com/vertex-ai/generative-ai/docs/thinking (accessed 2026-06-25)
89
temperature = true
910
tool_call = true
1011
structured_output = true

‎providers/aihubmix/provider.toml‎

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,9 @@
11
name = "AIHubMix"
22
npm = "@aihubmix/ai-sdk-provider"
3+
# Raw Chat: $.reasoning_effort = "none"|"minimal"|"low"|"medium"|"high"|"xhigh"; aliases are $.reasoning.effort and integer $.reasoning.max_tokens. "none" disables models that permit it. https://docs.aihubmix.com/cn/api/unified-inference (accessed 2026-06-25)
4+
# Raw Responses: $.reasoning.effort carries effort; this endpoint has no reasoning-token budget field. https://docs.aihubmix.com/cn/api-reference/openai-compatible/create-a-model-response (accessed 2026-06-25)
5+
# Raw Messages: $.thinking.type = "enabled"|"disabled"|"adaptive"; enabled uses $.thinking.budget_tokens >= 1024, and $.output_config.effort = "low"|"medium"|"high"|"xhigh"|"max" subject to model support. https://docs.aihubmix.com/cn/api-reference/anthropic-compatible/create-a-message (accessed 2026-06-25)
6+
# Raw Gemini native: $.generationConfig.thinkingConfig uses integer thinkingBudget (-1 dynamic; 0 off where supported) or string thinkingLevel; model bounds differ below. https://docs.aihubmix.com/cn/api-reference/google-vertex-ai-compatible/generate-content (accessed 2026-06-25)
37
env = ["AIHUBMIX_API_KEY"]
48
doc = "https://docs.aihubmix.com"
59

0 commit comments

Comments
 (0)