geoffsee/open-gsio
1
0
Fork 0
mirror of https://github.com/geoffsee/open-gsio.git synced 2025-09-08 22:56:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

improves interoperability of model providers, local and remote providers can be used together seemlessly

Browse Source
This commit is contained in:
geoffsee
2025-06-09 13:54:37 -04:00
committed by Geoff Seemueller
parent ad7dc5c0a6
commit f29bb6779c
14 changed files with 526 additions and 234 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
packages/server/ServerCoordinator.ts
View File

@@ -1,6 +1,9 @@
import { DurableObject } from "cloudflare:workers";
import {ProviderRepository} from "./providers/_ProviderRepository";
export default class ServerCoordinator extends DurableObject {
env;
state;
constructor(state, env) {
super(state, env);
this.state = state;
@@ -8,20 +11,24 @@ export default class ServerCoordinator extends DurableObject {
}
// Public method to calculate dynamic max tokens
async dynamicMaxTokens(input, maxOuputTokens) {
return 2000;
// const baseTokenLimit = 1024;
//
//
// const { encode } = await import("gpt-tokenizer/esm/model/gpt-4o");
//
// const inputTokens = Array.isArray(input)
// ? encode(input.map(i => i.content).join(' '))
// : encode(input);
//
// const scalingFactor = inputTokens.length > 300 ? 1.5 : 1;
//
// return Math.min(baseTokenLimit + Math.floor(inputTokens.length * scalingFactor^2), maxOuputTokens);
async dynamicMaxTokens(model, input, maxOuputTokens) {
const modelMeta = ProviderRepository.getModelMeta(model, this.env);
// The token‑limit information is stored in three different keys:
// max_completion_tokens
// context_window
// context_length
if('max_completion_tokens' in modelMeta) {
return modelMeta.max_completion_tokens;
} else if('context_window' in modelMeta) {
return modelMeta.context_window;
} else if('context_length' in modelMeta) {
return modelMeta.context_length;
} else {
return 8096;
}
}
// Public method to retrieve conversation history