🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.
TypeScript analyticsautogenevaluationlangchainlarge-language-modelsllama-indexllmllm-evaluationllm-observabilityllmopsmonitoringobservabilityopen-sourceopenaiplaygroundprompt-engineeringprompt-managementself-hostedycombinator
import {
PostModelsV1Body,
PostModelsV1Response,
} from "@/src/features/public-api/types/models";
import { createModelForApi } from "@/src/features/models/server/publicApiModelService";
import { defineTool } from "../../../core/define-tool";
import { buildModelUrl } from "@langfuse/shared/src/server";
import { runMcpTool } from "../../../core/run-mcp-tool";
import { z } from "zod";
const CreateModelBaseSchema = z.object({
modelName: z.string(),
matchPattern: z.string(),
startDate: z.string().optional(),
inputPrice: z.number().nonnegative().optional(),
outputPrice: z.number().nonnegative().optional(),
totalPrice: z.number().nonnegative().optional(),
unit: z.enum([
"TOKENS",
"CHARACTERS",
"MILLISECONDS",
"SECONDS",
"REQUESTS",
"IMAGES",
]),
tokenizerId: z.enum(["openai", "claude"]).optional(),
tokenizerConfig: z.any().optional(),
pricingTiers: z.array(z.any()).optional(),
});
export const [createModelTool, handleCreateModel] = defineTool({
name: "createModel",
description:
"Create a custom model definition for cost tracking/tokenization in the current project.",
action: "models:CUD",
baseSchema: CreateModelBaseSchema,
inputSchema: PostModelsV1Body,
handler: async (input, context) =>
runMcpTool({
spanName: "mcp.models.create",
context,
attributes: { "mcp.model_name": input.modelName },
fn: async () => {
const result = await createModelForApi({
projectId: context.projectId,
input,
auditScope: context,
});
const model = PostModelsV1Response.parse(result);
return {
...model,
url: buildModelUrl({
projectId: context.projectId,
modelId: model.id,
}),
};
},
}),
});