char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Aleksander Grygierui: Services consolidation refactor (#27239)3dc7285b4

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
1.1 KiB41 linesraw
1export const API_MODELS = {
2	LIST: '/v1/models',
3	LOAD: '/models/load',
4	SSE: '/models/sse',
5	UNLOAD: '/models/unload'
6};
7
8// chat completion routes, the control route drives realtime inference (e.g. end reasoning)
9export const API_CHAT = {
10	COMPLETIONS: './v1/chat/completions',
11	CONTROL: './v1/chat/completions/control'
12};
13
14// slot introspection, requires the --slots flag on the server
15export const API_SLOTS = {
16	LIST: './slots'
17};
18
19export const API_TOOLS = {
20	EXECUTE: '/tools',
21	LIST: '/tools'
22};
23
24// resumable stream routes, the conv::model identity travels as the conv_id query param
25// because model names can contain slashes that a path segment cannot carry
26// resume retry cadence while the owning model is still loading (server answers 503)
27export const STREAM_RESUME_RETRY_MS = 2000;
28
29export const API_STREAM = {
30	BASE: './v1/stream',
31	LOOKUP: './v1/streams/lookup'
32};
33
34// query params for the resumable stream routes
35export const STREAM_QUERY_PARAMS = {
36	CONV_ID: 'conv_id',
37	FROM: 'from'
38} as const;
39
40/** CORS proxy endpoint path */
41export const CORS_PROXY_ENDPOINT = '/cors-proxy';