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

add hardcoded context windows for Groq models

Browse source
This commit is contained in:
Russell Ballestrini 2025-07-12 07:28:13 -04:00
parent fd71442331
commit a864366c80
1 changed files with 34 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

41
src/models.js
View file

@ -84,13 +84,40 @@ export async function fetchModelsFromEndpoints() {
const models = jsonResponse.data || [];
console.log(`✅ Fetched ${models.length} models from ${endpoint.url}:`, models.map(m => m.id));
// Map each model to include its endpoint ID, unique ID, and model name
return models.map((model) => ({
...model,
modelName: model.id, // Explicitly store model name
endpointId: endpoint.id,
uniqueId: `${endpoint.id}-${model.id}`, // Unique ID with endpoint ID first
maxTokens: model.max_tokens || model.context_length || model.max_context_length || model.max_model_len || 8192, // Capture max tokens
}));
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;
// Hardcoded context windows for known Groq models (since API doesn't return them)
if (endpoint.id.includes('groq.com')) {
const groqContextWindows = {
'llama-3.1-8b-instant': 131072,
'llama-3.3-70b-versatile': 131072,
'llama-3.1-70b-versatile': 131072,
'llama3-8b-8192': 8192,
'llama3-70b-8192': 8192,
'mixtral-8x7b-32768': 32768,
'gemma2-9b-it': 8192,
'meta-llama/llama-4-scout-17b-16e-instruct': 131072,
'meta-llama/llama-4-maverick-17b-128e-instruct': 131072,
'deepseek-r1-distill-llama-70b': 131072,
'qwen/qwen3-32b': 131072,
'mistral-saba-24b': 32768,
'compound-beta': 131072,
'compound-beta-mini': 131072,
};
maxTokens = groqContextWindows[model.id] || 8192;
console.log(`Using hardcoded context window for Groq model ${model.id}: ${maxTokens} tokens`);
}
return {
...model,
modelName: model.id,
endpointId: endpoint.id,
uniqueId: `${endpoint.id}-${model.id}`,
maxTokens: maxTokens || 8192,
};
});
} catch (error) {
console.error(`Error fetching models from ${endpoint.url}:`, error);
return [];