engineering/uncloseai.com
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

move ollama model context windows to config.js endpoint configuration

Browse source
This commit is contained in:
Russell Ballestrini 2025-10-11 09:37:27 -04:00
parent 39db3f1004
commit 5fc300ecc2
2 changed files with 15 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

9
src/config.js
View file

@ -68,7 +68,14 @@ export async function getSystemMessage() {
export const VLLM_ENDPOINTS = [
{ id: "hermes.ai.unturf.com", url: "https://hermes.ai.unturf.com/v1" },
{ id: "hermes2.ai.unturf.com", url: "https://hermes2.ai.unturf.com/v1" },
{ id: "qwen.ai.unturf.com", url: "https://qwen.ai.unturf.com/v1" },
{
id: "qwen.ai.unturf.com",
url: "https://qwen.ai.unturf.com/v1",
// Ollama endpoints don't return max_model_len, so we specify it here
modelContextWindows: {
'hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q4_K_M': 32768,
}
},
];
// Function to get API configuration (custom or default)

8
src/models.js
View file

@ -87,7 +87,13 @@ export async function fetchModelsFromEndpoints() {
return models.map((model) => {
// Get context window size from API or use known values for Groq models
let maxTokens = model.max_tokens || model.context_length || model.max_context_length || model.max_model_len;
// Check if endpoint has modelContextWindows config (for Ollama/other endpoints without API max_model_len)
if (endpoint.modelContextWindows && endpoint.modelContextWindows[model.id]) {
maxTokens = endpoint.modelContextWindows[model.id];
console.log(`Using configured context window for model ${model.id}: ${maxTokens} tokens`);
}
// Hardcoded context windows for known Groq models (since API doesn't return them)
if (endpoint.id.includes('groq.com')) {
const groqContextWindows = {