feat: add Corti provider

providers/corti/models/corti-s1-mini.toml

@@ -0,0 +1,26 @@
+name = "Corti S1 Mini"
+description = "Efficient model for latency-sensitive generation, high-throughput workloads, and image analysis"
+attachment = true
+reasoning = true
+reasoning_options = []
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2026-06-13"
+last_updated = "2026-09-03"
+open_weights = false
+
+interleaved = true
+
+[cost]
+input = 1
+output = 4
+cache_read = 0.1
+
+[limit]
+context = 262_144
+output = 16_384
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/corti/models/corti-s1.toml

@@ -0,0 +1,29 @@
+name = "Corti S1"
+description = "Frontier reasoning model for complex multi-step reasoning, agentic coding, and deep analysis"
+attachment = false
+reasoning = true
+tool_call = true
+structured_output = true
+temperature = true
+release_date = "2026-06-13"
+last_updated = "2026-09-03"
+open_weights = false
+
+interleaved = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["high", "max"]
+
+[cost]
+input = 2
+output = 8
+cache_read = 0.2
+
+[limit]
+context = 262_144
+output = 16_384
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/corti/provider.toml

@@ -0,0 +1,13 @@
+name = "Corti"
+env = ["CORTI_BEARER"]
+npm = "@ai-sdk/openai-compatible"
+# Chat Completions: POST /v1/chat/completions; reasoning is always-on for
+# corti-s1 and corti-s1-mini (returned in `reasoning` field); -instant
+# variants produce no reasoning. Effort via `reasoning_effort = high|max`
+# (corti-s1 only; mini has no effort control).
+# Responses: POST /v1/responses; reasoning as `type: "reasoning"` output items.
+# Anthropic Messages: POST /anthropic/v1/messages; reasoning not exposed.
+# https://docs.corti.ai/models/models (accessed 2026-09-03)
+# https://ai.eu.corti.app/v1/models (accessed 2026-09-03)
+api = "https://ai.eu.corti.app/v1"
+doc = "https://docs.corti.ai"