zack3d/OAI-Proxy
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

o3-pro

Browse Source
This commit is contained in:
reanon
2025-06-11 10:47:48 +02:00
parent 0102c7a6a5
commit 38e2980419
8 changed files with 19 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CODEBASE_GUIDE.md
+2 -2
View File
@@ -131,7 +131,7 @@ This project provides a proxy layer for various Large Language Models (LLMs) and
### Supported Models & Providers
* **OpenAI:** Handled in `src/proxy/openai.ts`. Supports models like GPT-4, GPT-3.5-turbo, as well as o-series models (o1, o1-mini, o1-pro, o3, o3-mini, o4-mini). Handles chat completions and potentially image generation (`src/proxy/openai-image.ts`).
* **OpenAI:** Handled in `src/proxy/openai.ts`. Supports models like GPT-4, GPT-3.5-turbo, as well as o-series models (o1, o1-mini, o1-pro, o3, o3-mini, o3-pro, o4-mini). Handles chat completions and potentially image generation (`src/proxy/openai-image.ts`).
* **Anthropic:** Handled in `src/proxy/anthropic.ts`. Supports Claude models. May use AWS Bedrock (`src/proxy/aws-claude.ts`) or Anthropic's direct API.
* **Google AI / Vertex AI:** Handled in `src/proxy/google-ai.ts` and `src/proxy/gcp.ts`. Supports Gemini models (gemini-flash, gemini-pro, gemini-ultra).
* **Mistral AI:** Handled in `src/proxy/mistral-ai.ts`. Supports Mistral models via their API or potentially AWS (`src/proxy/aws-mistral.ts`).
@@ -229,7 +229,7 @@ The codebase handles several patterns for model naming and versioning:
1. **Date-stamped Models:** Many models include date stamps (e.g., `gpt-4-0125-preview`). The regex patterns in `OPENAI_MODEL_FAMILY_MAP` account for these with patterns like `^gpt-4o(-\\d{4}-\\d{2}-\\d{2})?$`.
2. **O-Series Models:** OpenAI's o-series models (o1, o1-mini, o1-pro, o3, o3-mini, o4-mini) follow a different naming convention. The codebase handles these with dedicated model families and regex patterns.
2. **O-Series Models:** OpenAI's o-series models (o1, o1-mini, o1-pro, o3, o3-mini, o3-pro, o4-mini) follow a different naming convention. The codebase handles these with dedicated model families and regex patterns.
3. **Preview/Non-Preview Variants:** Some models have preview variants (e.g., `gpt-4.5-preview`). The regex patterns in `OPENAI_MODEL_FAMILY_MAP` account for these with patterns like `^gpt-4\\.5(-preview)?(-\\d{4}-\\d{2}-\\d{2})?$`.
src/config.ts
+1 -1
View File
@@ -880,6 +880,6 @@ function parseCsv(val: string): string[] {
function getDefaultModelFamilies(): ModelFamily[] {
return MODEL_FAMILIES.filter(
(f) => !f.includes("o1-pro")
(f) => !f.includes("o1-pro") && !f.includes("o3-pro")
) as ModelFamily[];
}
src/info-page.ts
+2
View File
@@ -43,6 +43,7 @@ const MODEL_FAMILY_FRIENDLY_NAME: { [f in ModelFamily]: string } = {
o1: "OpenAI o1",
"o1-mini": "OpenAI o1 mini",
"o1-pro": "OpenAI o1 pro",
"o3-pro": "OpenAI o3 pro",
"o3-mini": "OpenAI o3 mini",
"o3": "OpenAI o3",
"o4-mini": "OpenAI o4 mini",
@@ -78,6 +79,7 @@ const MODEL_FAMILY_FRIENDLY_NAME: { [f in ModelFamily]: string } = {
"azure-o1": "Azure o1",
"azure-o1-mini": "Azure o1 mini",
"azure-o1-pro": "Azure o1 pro",
"azure-o3-pro": "Azure o3 pro",
"azure-o3-mini": "Azure o3 mini",
"azure-o3": "Azure o3",
"azure-o4-mini": "Azure o4 mini",
src/proxy/middleware/request/preprocessors/validate-context-size.ts
+2
View File
@@ -91,6 +91,8 @@ export const validateContextSize: RequestPreprocessor = async (req) => {
modelMax = 128000;
} else if (model.match(/^o1-pro(-\d{4}-\d{2}-\d{2})?$/)) {
modelMax = 200000;
} else if (model.match(/^o3-pro(-\d{4}-\d{2}-\d{2})?$/)) {
modelMax = 200000;
} else if (model.match(/^o1-preview(-\d{4}-\d{2}-\d{2})?$/)) {
modelMax = 128000;
} else if (model.match(/gpt-3.5-turbo/)) {
src/proxy/openai.ts
+1
View File
@@ -357,6 +357,7 @@ const setupChunkedTransfer: RequestHandler = (req, res, next) => {
// Functions to handle model-specific API routing
function shouldUseResponsesApi(model: string): boolean {
return model === "o1-pro" || model.startsWith("o1-pro-") ||
model === "o3-pro" || model.startsWith("o3-pro-") ||
model === "codex-mini-latest" || model.startsWith("codex-mini-");
}
src/shared/key-management/index.ts
+1 -1
View File
@@ -6,7 +6,7 @@ export type APIFormat =
| "openai"
| "openai-text"
| "openai-image"
| "openai-responses" // New OpenAI Responses API for o1-pro model
| "openai-responses" // OpenAI Responses API (e.g., for o1-pro, o3-pro)
| "anthropic-chat" // Anthropic's newer messages array format
| "anthropic-text" // Legacy flat string prompt format
| "google-ai"
src/shared/models.ts
+6
View File
@@ -33,6 +33,7 @@ export type OpenAIModelFamily =
| "o1"
| "o1-mini"
| "o1-pro"
| "o3-pro"
| "o3-mini"
| "o3"
| "o4-mini"
@@ -90,6 +91,7 @@ export const MODEL_FAMILIES = (<A extends readonly ModelFamily[]>(
"o1",
"o1-mini",
"o1-pro",
"o3-pro",
"o3-mini",
"o3",
"o4-mini",
@@ -126,6 +128,7 @@ export const MODEL_FAMILIES = (<A extends readonly ModelFamily[]>(
"azure-o1",
"azure-o1-mini",
"azure-o1-pro",
"azure-o3-pro",
"azure-o3-mini",
"azure-o3",
"azure-o4-mini",
@@ -168,6 +171,7 @@ export const MODEL_FAMILY_SERVICE: {
"o1": "openai",
"o1-mini": "openai",
"o1-pro": "openai",
"o3-pro": "openai",
"o3-mini": "openai",
"o3": "openai",
"o4-mini": "openai",
@@ -197,6 +201,7 @@ export const MODEL_FAMILY_SERVICE: {
"azure-o1": "azure",
"azure-o1-mini": "azure",
"azure-o1-pro": "azure",
"azure-o3-pro": "azure",
"azure-o3-mini": "azure",
"azure-o3": "azure",
"azure-o4-mini": "azure",
@@ -235,6 +240,7 @@ export const OPENAI_MODEL_FAMILY_MAP: { [regex: string]: OpenAIModelFamily } = {
"^dall-e-\\d{1}$": "dall-e",
"^o1-mini(-\\d{4}-\\d{2}-\\d{2})?$": "o1-mini",
"^o1-pro(-\\d{4}-\\d{2}-\\d{2})?$": "o1-pro",
"^o3-pro(-\\d{4}-\\d{2}-\\d{2})?$": "o3-pro",
"^o1(-\\d{4}-\\d{2}-\\d{2})?$": "o1",
"^o3-mini(-\\d{4}-\\d{2}-\\d{2})?$": "o3-mini",
"^o3(-\\d{4}-\\d{2}-\\d{2})?$": "o3",
src/shared/stats.ts
+3 -1
View File
@@ -19,13 +19,15 @@ const MODEL_PRICING: Record<ModelFamily, { input: number; output: number } | und
"azure-gpt4-turbo": { input: 10.00, output: 30.00 },
"o1-pro": { input: 150.00, output: 600.00 },
"azure-o1-pro": { input: 150.00, output: 600.00 },
"o3-pro": { input: 20.00, output: 80.00 },
"azure-o3-pro": { input: 20.00, output: 80.00 },
"o1": { input: 15.00, output: 60.00 },
"azure-o1": { input: 15.00, output: 60.00 },
"o1-mini": { input: 1.10, output: 4.40 },
"azure-o1-mini": { input: 1.10, output: 4.40 },
"o3-mini": { input: 1.10, output: 4.40 },
"azure-o3-mini": { input: 1.10, output: 4.40 },
"o3": { input: 2.00, output: 8.00 },
"o3": { input: 10.00, output: 40.00 },
"azure-o3": { input: 10.00, output: 40.00 },
"o4-mini": { input: 1.10, output: 4.40 },
"azure-o4-mini": { input: 1.10, output: 4.40 },