kunthawat/moreminimore-vibe
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Simplify provider logic and migrate getContextWindow (#142)

Browse Source
This commit is contained in:
Will Chen
2025-05-12 22:18:49 -07:00
committed by GitHub
parent 11ba46db38
commit 877c8f7f4f
7 changed files with 48 additions and 67 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
src/ipc/utils/get_model_client.ts
View File

@@ -5,7 +5,7 @@ import { createOpenRouter } from "@openrouter/ai-sdk-provider";
import { createOllama } from "ollama-ai-provider";
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import type { LargeLanguageModel, UserSettings } from "../../lib/schemas";
import { AUTO_MODELS, MODEL_OPTIONS } from "../../constants/models";
import { AUTO_MODELS } from "../../constants/models";
import { getEnvVar } from "./read_env";
import log from "electron-log";
import { getLanguageModelProviders } from "../shared/language_model_helpers";
@@ -140,19 +140,3 @@ export async function getModelClient(
}
}
}
// Most models support at least 8000 output tokens so we use it as a default value.
const DEFAULT_MAX_TOKENS = 8_000;
export function getMaxTokens(model: LargeLanguageModel) {
if (!MODEL_OPTIONS[model.provider as keyof typeof MODEL_OPTIONS]) {
logger.warn(
`Model provider ${model.provider} not found in MODEL_OPTIONS. Using default max tokens.`,
);
return DEFAULT_MAX_TOKENS;
}
const modelOption = MODEL_OPTIONS[
model.provider as keyof typeof MODEL_OPTIONS
].find((m) => m.name === model.name);
return modelOption?.maxOutputTokens || DEFAULT_MAX_TOKENS;
}

34
src/ipc/utils/token_utils.ts
View File

@@ -1,9 +1,8 @@
import { LargeLanguageModel } from "@/lib/schemas";
import { readSettings } from "../../main/settings";
import { Message } from "../ipc_types";
import { MODEL_OPTIONS } from "../../constants/models";
import log from "electron-log";
const logger = log.scope("token_utils");
import { getLanguageModels } from "../shared/language_model_helpers";
// Estimate tokens (4 characters per token)
export const estimateTokens = (text: string): number => {
@@ -19,17 +18,26 @@ export const estimateMessagesTokens = (messages: Message[]): number => {
const DEFAULT_CONTEXT_WINDOW = 128_000;
export function getContextWindow() {
export async function getContextWindow() {
const settings = readSettings();
const model = settings.selectedModel;
if (!MODEL_OPTIONS[model.provider as keyof typeof MODEL_OPTIONS]) {
logger.warn(
`Model provider ${model.provider} not found in MODEL_OPTIONS. Using default max tokens.`,
);
return DEFAULT_CONTEXT_WINDOW;
}
const modelOption = MODEL_OPTIONS[
model.provider as keyof typeof MODEL_OPTIONS
].find((m) => m.name === model.name);
const models = await getLanguageModels({
providerId: model.provider,
});
const modelOption = models.find((m) => m.apiName === model.name);
return modelOption?.contextWindow || DEFAULT_CONTEXT_WINDOW;
}
// Most models support at least 8000 output tokens so we use it as a default value.
const DEFAULT_MAX_TOKENS = 8_000;
export async function getMaxTokens(model: LargeLanguageModel) {
const models = await getLanguageModels({
providerId: model.provider,
});
const modelOption = models.find((m) => m.apiName === model.name);
return modelOption?.maxOutputTokens || DEFAULT_MAX_TOKENS;
}