@oh-my-pi/pi-ai 14.5.14 → 14.6.1

package/CHANGELOG.md

@@ -2,6 +2,34 @@
 
 ## [Unreleased]
 
+## [14.6.0] - 2026-05-02
+
+### Added
+
+- Added `disableReasoning` to stream and OpenAI completion options to force reasoning off for models that support it, sending `reasoning: { enabled: false }` for OpenRouter-compatible requests
+- Added `thinkingDisplay` option to Anthropic options to control whether adaptive and explicit reasoning is returned as `summarized` or `omitted`
+- Added Anthropic model compatibility flags `supportsEagerToolInputStreaming` and `supportsLongCacheRetention` for API-capability-specific request behavior
+
+### Changed
+
+- Changed Anthropic request payloads to send `thinking: { type: "disabled" }` when `thinkingEnabled` is explicitly `false` on reasoning-enabled models
+- Changed Anthropic cache retention handling so `cacheRetention: "long"` now uses `ttl: "1h"` only for canonical Anthropic endpoints with long-cache support
+- Changed Anthropic tool schema generation to include `eager_input_streaming` only on models that advertise support
+- Changed Anthropic OAuth login flow to include browser fallback guidance and richer error context when token exchange or refresh fails
+
+### Fixed
+
+- Fixed Anthropic non-thinking requests to include the caller-provided `temperature` value in request payloads
+- Fixed Anthropic `claude-opus-4-7` non-thinking payloads to omit sampling fields (`temperature`, `top_p`, and `top_k`)
+- Fixed OpenAI Codex base URL normalization so configured base URLs with or without `/codex` or `/codex/responses` now resolve to `/codex/responses`
+- Fixed OpenAI Codex websocket handling to parse JSON from non-string message payloads including `ArrayBuffer`, typed arrays, and `Blob` values
+- Fixed OpenAI Codex websocket handshakes to replace stale `openai-beta` values with the websocket beta and avoid sending request-body headers over websocket transport
+- Fixed abort tracking so caller-initiated cancellations are treated as user aborts even after local watchdog timeouts, preventing unintended automatic retries
+- Fixed Anthropic stream handling to parse raw SSE envelopes directly, ignore unrelated events, and repair malformed JSON in SSE payloads
+- Fixed Anthropic streaming to emit an explicit error when the SSE stream ends without a `message_stop` event
+- Fixed OpenAI Codex websocket continuations to send true `previous_response_id` deltas for `store: false` transcripts, expose request stats, and default text verbosity to `low` unless explicitly overridden.
+- Fixed OpenAI Codex websocket append reuse after `response.completed` terminal events.
+
 ## [14.5.14] - 2026-05-01
 ### Added
 
@@ -2208,4 +2236,4 @@ _Dedicated to Peter's shoulder ([@steipete](https://twitter.com/steipete))_
 
 ## [0.9.4] - 2025-11-26
 
-Initial release with multi-provider LLM support.
+Initial release with multi-provider LLM support.

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-ai",
-	"version": "14.5.14",
+	"version": "14.6.1",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"homepage": "https://github.com/can1357/oh-my-pi",
 	"author": "Can Boluk",
@@ -46,8 +46,8 @@
 		"@aws-sdk/credential-provider-node": "^3.972.36",
 		"@bufbuild/protobuf": "^2.12.0",
 		"@google/genai": "^1.50.1",
-		"@oh-my-pi/pi-natives": "14.5.14",
-		"@oh-my-pi/pi-utils": "14.5.14",
+		"@oh-my-pi/pi-natives": "14.6.1",
+		"@oh-my-pi/pi-utils": "14.6.1",
 		"@sinclair/typebox": "^0.34.49",
 		"@smithy/node-http-handler": "^4.6.1",
 		"ajv": "^8.20.0",

package/src/index.ts

@@ -18,7 +18,7 @@ export * from "./providers/google-gemini-headers";
 export type * from "./providers/google-vertex";
 export * from "./providers/kimi";
 export * from "./providers/ollama";
-export type { OpenAICodexResponsesOptions } from "./providers/openai-codex-responses";
+export * from "./providers/openai-codex-responses";
 export * from "./providers/openai-completions";
 export * from "./providers/openai-responses";
 export * from "./providers/synthetic";

package/src/models.json

@@ -28105,6 +28105,44 @@
 			"contextWindow": 222222,
 			"maxTokens": 8888
 		},
+		"poolside/laguna-m.1": {
+			"id": "poolside/laguna-m.1",
+			"name": "poolside/laguna-m.1",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
+		"poolside/laguna-xs.2": {
+			"id": "poolside/laguna-xs.2",
+			"name": "poolside/laguna-xs.2",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
 		"qvq-max": {
 			"id": "qvq-max",
 			"name": "qvq-max",
@@ -35776,9 +35814,9 @@
 		"minimax-m2.7": {
 			"id": "minimax-m2.7",
 			"name": "MiniMax M2.7",
-			"api": "anthropic-messages",
+			"api": "openai-completions",
 			"provider": "opencode-go",
-			"baseUrl": "https://opencode.ai/zen/go",
+			"baseUrl": "https://opencode.ai/zen/go/v1",
 			"reasoning": true,
 			"input": [
 				"text"
@@ -35792,7 +35830,7 @@
 			"contextWindow": 204800,
 			"maxTokens": 131072,
 			"thinking": {
-				"mode": "budget",
+				"mode": "effort",
 				"minLevel": "minimal",
 				"maxLevel": "xhigh"
 			}
@@ -38485,7 +38523,7 @@
 				"cacheRead": 0.024999999999999998,
 				"cacheWrite": 0.08333333333333334
 			},
-			"contextWindow": 1000000,
+			"contextWindow": 1048576,
 			"maxTokens": 8192
 		},
 		"google/gemini-2.0-flash-lite-001": {
@@ -42996,13 +43034,13 @@
 				"image"
 			],
 			"cost": {
-				"input": 0.325,
-				"output": 3.25,
+				"input": 0.32,
+				"output": 3.1999999999999997,
 				"cacheRead": 0,
 				"cacheWrite": 0
 			},
-			"contextWindow": 256000,
-			"maxTokens": 65536,
+			"contextWindow": 262144,
+			"maxTokens": 81920,
 			"thinking": {
 				"mode": "effort",
 				"minLevel": "minimal",
@@ -55034,6 +55072,26 @@
 			"contextWindow": 2000000,
 			"maxTokens": 30000
 		},
+		"x-ai/grok-4.3": {
+			"id": "x-ai/grok-4.3",
+			"name": "xAI: Grok 4.3",
+			"api": "openai-completions",
+			"provider": "zenmux",
+			"baseUrl": "https://zenmux.ai/api/v1",
+			"reasoning": false,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 1.25,
+				"output": 2.5,
+				"cacheRead": 0.2,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1000000,
+			"maxTokens": 8888
+		},
 		"x-ai/grok-code-fast-1": {
 			"id": "x-ai/grok-code-fast-1",
 			"name": "Grok Code Fast 1",

package/src/provider-models/openai-compat.ts

@@ -1868,12 +1868,18 @@ function createOpenCodeApiResolution(
 }
 
 const OPENCODE_ZEN_API_RESOLUTION = createOpenCodeApiResolution("https://opencode.ai/zen");
-// OpenCode Go: models.dev declares qwen3.5-plus / qwen3.6-plus with
-// `provider.npm = "@ai-sdk/anthropic"`, but per the OpenCode Go endpoint table
-// (https://opencode.ai/docs/go/#endpoints) they are served via @ai-sdk/alibaba
-// at https://opencode.ai/zen/go/v1/chat/completions (OpenAI-compatible).
-// Override the resolver so regenerating models.json keeps the correct routing.
+// OpenCode Go: models.dev declares minimax-m2.7 / qwen3.5-plus / qwen3.6-plus
+// with `provider.npm = "@ai-sdk/anthropic"`, but the OpenCode Go gateway only
+// serves them at `https://opencode.ai/zen/go/v1/chat/completions` (verified
+// against https://opencode.ai/zen/go/v1/models and the upstream endpoint
+// table at https://opencode.ai/docs/go/#endpoints — minimax-m2.5 works the
+// same way and lacks an `npm` field on models.dev so it already falls through
+// to the openai-completions default). Without this override the resolver
+// would POST anthropic-style requests to /v1/messages and the gateway would
+// return its `Page Not Found` HTML (issue #887). Override the resolver so
+// regenerating models.json keeps the correct routing.
 const OPENCODE_GO_API_RESOLUTION = createOpenCodeApiResolution("https://opencode.ai/zen/go", {
+	"minimax-m2.7": "openai-completions",
 	"qwen3.5-plus": "openai-completions",
 	"qwen3.6-plus": "openai-completions",
 });