inference-server 1.0.0-beta.19
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +216 -0
- package/dist/api/openai/enums.d.ts +4 -0
- package/dist/api/openai/enums.js +17 -0
- package/dist/api/openai/enums.js.map +1 -0
- package/dist/api/openai/handlers/chat.d.ts +3 -0
- package/dist/api/openai/handlers/chat.js +358 -0
- package/dist/api/openai/handlers/chat.js.map +1 -0
- package/dist/api/openai/handlers/completions.d.ts +3 -0
- package/dist/api/openai/handlers/completions.js +169 -0
- package/dist/api/openai/handlers/completions.js.map +1 -0
- package/dist/api/openai/handlers/embeddings.d.ts +3 -0
- package/dist/api/openai/handlers/embeddings.js +74 -0
- package/dist/api/openai/handlers/embeddings.js.map +1 -0
- package/dist/api/openai/handlers/images.d.ts +0 -0
- package/dist/api/openai/handlers/images.js +4 -0
- package/dist/api/openai/handlers/images.js.map +1 -0
- package/dist/api/openai/handlers/models.d.ts +3 -0
- package/dist/api/openai/handlers/models.js +23 -0
- package/dist/api/openai/handlers/models.js.map +1 -0
- package/dist/api/openai/handlers/transcription.d.ts +0 -0
- package/dist/api/openai/handlers/transcription.js +4 -0
- package/dist/api/openai/handlers/transcription.js.map +1 -0
- package/dist/api/openai/index.d.ts +7 -0
- package/dist/api/openai/index.js +14 -0
- package/dist/api/openai/index.js.map +1 -0
- package/dist/api/parseJSONRequestBody.d.ts +2 -0
- package/dist/api/parseJSONRequestBody.js +24 -0
- package/dist/api/parseJSONRequestBody.js.map +1 -0
- package/dist/api/v1/index.d.ts +2 -0
- package/dist/api/v1/index.js +29 -0
- package/dist/api/v1/index.js.map +1 -0
- package/dist/cli.d.ts +1 -0
- package/dist/cli.js +10 -0
- package/dist/cli.js.map +1 -0
- package/dist/engines/gpt4all/engine.d.ts +34 -0
- package/dist/engines/gpt4all/engine.js +357 -0
- package/dist/engines/gpt4all/engine.js.map +1 -0
- package/dist/engines/gpt4all/util.d.ts +3 -0
- package/dist/engines/gpt4all/util.js +29 -0
- package/dist/engines/gpt4all/util.js.map +1 -0
- package/dist/engines/index.d.ts +19 -0
- package/dist/engines/index.js +21 -0
- package/dist/engines/index.js.map +1 -0
- package/dist/engines/node-llama-cpp/engine.d.ts +49 -0
- package/dist/engines/node-llama-cpp/engine.js +666 -0
- package/dist/engines/node-llama-cpp/engine.js.map +1 -0
- package/dist/engines/node-llama-cpp/types.d.ts +13 -0
- package/dist/engines/node-llama-cpp/types.js +2 -0
- package/dist/engines/node-llama-cpp/types.js.map +1 -0
- package/dist/engines/node-llama-cpp/util.d.ts +15 -0
- package/dist/engines/node-llama-cpp/util.js +84 -0
- package/dist/engines/node-llama-cpp/util.js.map +1 -0
- package/dist/engines/node-llama-cpp/validateModelFile.d.ts +8 -0
- package/dist/engines/node-llama-cpp/validateModelFile.js +36 -0
- package/dist/engines/node-llama-cpp/validateModelFile.js.map +1 -0
- package/dist/engines/stable-diffusion-cpp/engine.d.ts +90 -0
- package/dist/engines/stable-diffusion-cpp/engine.js +294 -0
- package/dist/engines/stable-diffusion-cpp/engine.js.map +1 -0
- package/dist/engines/stable-diffusion-cpp/types.d.ts +3 -0
- package/dist/engines/stable-diffusion-cpp/types.js +2 -0
- package/dist/engines/stable-diffusion-cpp/types.js.map +1 -0
- package/dist/engines/stable-diffusion-cpp/util.d.ts +4 -0
- package/dist/engines/stable-diffusion-cpp/util.js +55 -0
- package/dist/engines/stable-diffusion-cpp/util.js.map +1 -0
- package/dist/engines/stable-diffusion-cpp/validateModelFiles.d.ts +19 -0
- package/dist/engines/stable-diffusion-cpp/validateModelFiles.js +91 -0
- package/dist/engines/stable-diffusion-cpp/validateModelFiles.js.map +1 -0
- package/dist/engines/transformers-js/engine.d.ts +37 -0
- package/dist/engines/transformers-js/engine.js +538 -0
- package/dist/engines/transformers-js/engine.js.map +1 -0
- package/dist/engines/transformers-js/types.d.ts +7 -0
- package/dist/engines/transformers-js/types.js +2 -0
- package/dist/engines/transformers-js/types.js.map +1 -0
- package/dist/engines/transformers-js/util.d.ts +7 -0
- package/dist/engines/transformers-js/util.js +36 -0
- package/dist/engines/transformers-js/util.js.map +1 -0
- package/dist/engines/transformers-js/validateModelFiles.d.ts +17 -0
- package/dist/engines/transformers-js/validateModelFiles.js +133 -0
- package/dist/engines/transformers-js/validateModelFiles.js.map +1 -0
- package/dist/experiments/ChatWithVision.d.ts +11 -0
- package/dist/experiments/ChatWithVision.js +91 -0
- package/dist/experiments/ChatWithVision.js.map +1 -0
- package/dist/experiments/StableDiffPromptGenerator.d.ts +0 -0
- package/dist/experiments/StableDiffPromptGenerator.js +4 -0
- package/dist/experiments/StableDiffPromptGenerator.js.map +1 -0
- package/dist/experiments/VoiceFunctionCall.d.ts +18 -0
- package/dist/experiments/VoiceFunctionCall.js +51 -0
- package/dist/experiments/VoiceFunctionCall.js.map +1 -0
- package/dist/http.d.ts +19 -0
- package/dist/http.js +54 -0
- package/dist/http.js.map +1 -0
- package/dist/index.d.ts +7 -0
- package/dist/index.js +8 -0
- package/dist/index.js.map +1 -0
- package/dist/instance.d.ts +88 -0
- package/dist/instance.js +594 -0
- package/dist/instance.js.map +1 -0
- package/dist/lib/acquireFileLock.d.ts +7 -0
- package/dist/lib/acquireFileLock.js +38 -0
- package/dist/lib/acquireFileLock.js.map +1 -0
- package/dist/lib/calculateContextIdentity.d.ts +7 -0
- package/dist/lib/calculateContextIdentity.js +39 -0
- package/dist/lib/calculateContextIdentity.js.map +1 -0
- package/dist/lib/calculateFileChecksum.d.ts +1 -0
- package/dist/lib/calculateFileChecksum.js +16 -0
- package/dist/lib/calculateFileChecksum.js.map +1 -0
- package/dist/lib/copyDirectory.d.ts +6 -0
- package/dist/lib/copyDirectory.js +27 -0
- package/dist/lib/copyDirectory.js.map +1 -0
- package/dist/lib/decodeAudio.d.ts +1 -0
- package/dist/lib/decodeAudio.js +26 -0
- package/dist/lib/decodeAudio.js.map +1 -0
- package/dist/lib/downloadModelFile.d.ts +10 -0
- package/dist/lib/downloadModelFile.js +58 -0
- package/dist/lib/downloadModelFile.js.map +1 -0
- package/dist/lib/flattenMessageTextContent.d.ts +2 -0
- package/dist/lib/flattenMessageTextContent.js +11 -0
- package/dist/lib/flattenMessageTextContent.js.map +1 -0
- package/dist/lib/getCacheDirPath.d.ts +12 -0
- package/dist/lib/getCacheDirPath.js +31 -0
- package/dist/lib/getCacheDirPath.js.map +1 -0
- package/dist/lib/loadImage.d.ts +12 -0
- package/dist/lib/loadImage.js +30 -0
- package/dist/lib/loadImage.js.map +1 -0
- package/dist/lib/logger.d.ts +12 -0
- package/dist/lib/logger.js +98 -0
- package/dist/lib/logger.js.map +1 -0
- package/dist/lib/math.d.ts +7 -0
- package/dist/lib/math.js +30 -0
- package/dist/lib/math.js.map +1 -0
- package/dist/lib/resolveModelFileLocation.d.ts +15 -0
- package/dist/lib/resolveModelFileLocation.js +41 -0
- package/dist/lib/resolveModelFileLocation.js.map +1 -0
- package/dist/lib/util.d.ts +7 -0
- package/dist/lib/util.js +61 -0
- package/dist/lib/util.js.map +1 -0
- package/dist/lib/validateModelFile.d.ts +9 -0
- package/dist/lib/validateModelFile.js +62 -0
- package/dist/lib/validateModelFile.js.map +1 -0
- package/dist/lib/validateModelOptions.d.ts +3 -0
- package/dist/lib/validateModelOptions.js +23 -0
- package/dist/lib/validateModelOptions.js.map +1 -0
- package/dist/pool.d.ts +61 -0
- package/dist/pool.js +512 -0
- package/dist/pool.js.map +1 -0
- package/dist/server.d.ts +59 -0
- package/dist/server.js +221 -0
- package/dist/server.js.map +1 -0
- package/dist/standalone.d.ts +1 -0
- package/dist/standalone.js +306 -0
- package/dist/standalone.js.map +1 -0
- package/dist/store.d.ts +60 -0
- package/dist/store.js +203 -0
- package/dist/store.js.map +1 -0
- package/dist/types/completions.d.ts +57 -0
- package/dist/types/completions.js +2 -0
- package/dist/types/completions.js.map +1 -0
- package/dist/types/index.d.ts +326 -0
- package/dist/types/index.js +2 -0
- package/dist/types/index.js.map +1 -0
- package/docs/engines.md +28 -0
- package/docs/gpu.md +72 -0
- package/docs/http-api.md +147 -0
- package/examples/all-options.js +108 -0
- package/examples/chat-cli.js +56 -0
- package/examples/chat-server.js +65 -0
- package/examples/concurrency.js +70 -0
- package/examples/express.js +70 -0
- package/examples/pool.js +91 -0
- package/package.json +113 -0
- package/src/api/openai/enums.ts +20 -0
- package/src/api/openai/handlers/chat.ts +408 -0
- package/src/api/openai/handlers/completions.ts +196 -0
- package/src/api/openai/handlers/embeddings.ts +92 -0
- package/src/api/openai/handlers/images.ts +3 -0
- package/src/api/openai/handlers/models.ts +33 -0
- package/src/api/openai/handlers/transcription.ts +2 -0
- package/src/api/openai/index.ts +16 -0
- package/src/api/parseJSONRequestBody.ts +26 -0
- package/src/api/v1/DRAFT.md +16 -0
- package/src/api/v1/index.ts +37 -0
- package/src/cli.ts +9 -0
- package/src/engines/gpt4all/engine.ts +441 -0
- package/src/engines/gpt4all/util.ts +31 -0
- package/src/engines/index.ts +28 -0
- package/src/engines/node-llama-cpp/engine.ts +811 -0
- package/src/engines/node-llama-cpp/types.ts +17 -0
- package/src/engines/node-llama-cpp/util.ts +126 -0
- package/src/engines/node-llama-cpp/validateModelFile.ts +46 -0
- package/src/engines/stable-diffusion-cpp/engine.ts +369 -0
- package/src/engines/stable-diffusion-cpp/types.ts +54 -0
- package/src/engines/stable-diffusion-cpp/util.ts +58 -0
- package/src/engines/stable-diffusion-cpp/validateModelFiles.ts +119 -0
- package/src/engines/transformers-js/engine.ts +659 -0
- package/src/engines/transformers-js/types.ts +25 -0
- package/src/engines/transformers-js/util.ts +40 -0
- package/src/engines/transformers-js/validateModelFiles.ts +168 -0
- package/src/experiments/ChatWithVision.ts +103 -0
- package/src/experiments/StableDiffPromptGenerator.ts +2 -0
- package/src/experiments/VoiceFunctionCall.ts +71 -0
- package/src/http.ts +72 -0
- package/src/index.ts +7 -0
- package/src/instance.ts +723 -0
- package/src/lib/acquireFileLock.ts +38 -0
- package/src/lib/calculateContextIdentity.ts +53 -0
- package/src/lib/calculateFileChecksum.ts +18 -0
- package/src/lib/copyDirectory.ts +29 -0
- package/src/lib/decodeAudio.ts +39 -0
- package/src/lib/downloadModelFile.ts +70 -0
- package/src/lib/flattenMessageTextContent.ts +19 -0
- package/src/lib/getCacheDirPath.ts +34 -0
- package/src/lib/loadImage.ts +46 -0
- package/src/lib/logger.ts +112 -0
- package/src/lib/math.ts +31 -0
- package/src/lib/resolveModelFileLocation.ts +49 -0
- package/src/lib/util.ts +75 -0
- package/src/lib/validateModelFile.ts +71 -0
- package/src/lib/validateModelOptions.ts +31 -0
- package/src/pool.ts +651 -0
- package/src/server.ts +270 -0
- package/src/standalone.ts +320 -0
- package/src/store.ts +278 -0
- package/src/types/completions.ts +86 -0
- package/src/types/index.ts +488 -0
- package/tsconfig.json +29 -0
- package/tsconfig.release.json +11 -0
- package/vitest.config.ts +18 -0
|
@@ -0,0 +1,169 @@
|
|
|
1
|
+
import { parseJSONRequestBody } from '../../../api/parseJSONRequestBody.js';
|
|
2
|
+
import { omitEmptyValues } from '../../../lib/util.js';
|
|
3
|
+
import { finishReasonMap } from '../enums.js';
|
|
4
|
+
// v1/completions
|
|
5
|
+
// https://platform.openai.com/docs/api-reference/completions/create
|
|
6
|
+
export function createCompletionHandler(modelServer) {
|
|
7
|
+
return async (req, res) => {
|
|
8
|
+
let args;
|
|
9
|
+
try {
|
|
10
|
+
const body = await parseJSONRequestBody(req);
|
|
11
|
+
args = body;
|
|
12
|
+
}
|
|
13
|
+
catch (e) {
|
|
14
|
+
console.error(e);
|
|
15
|
+
res.writeHead(400, { 'Content-Type': 'application/json' });
|
|
16
|
+
res.end(JSON.stringify({ error: 'Invalid request' }));
|
|
17
|
+
return;
|
|
18
|
+
}
|
|
19
|
+
// TODO ajv schema validation?
|
|
20
|
+
if (!args.model || !args.prompt) {
|
|
21
|
+
res.writeHead(400, { 'Content-Type': 'application/json' });
|
|
22
|
+
res.end(JSON.stringify({ error: 'Invalid request' }));
|
|
23
|
+
return;
|
|
24
|
+
}
|
|
25
|
+
if (!modelServer.modelExists(args.model)) {
|
|
26
|
+
res.writeHead(400, { 'Content-Type': 'application/json' });
|
|
27
|
+
res.end(JSON.stringify({ error: 'Invalid model' }));
|
|
28
|
+
return;
|
|
29
|
+
}
|
|
30
|
+
const controller = new AbortController();
|
|
31
|
+
req.on('close', () => {
|
|
32
|
+
console.debug('Client closed connection');
|
|
33
|
+
controller.abort();
|
|
34
|
+
});
|
|
35
|
+
req.on('end', () => {
|
|
36
|
+
console.debug('Client ended connection');
|
|
37
|
+
controller.abort();
|
|
38
|
+
});
|
|
39
|
+
try {
|
|
40
|
+
if (args.stream) {
|
|
41
|
+
res.writeHead(200, {
|
|
42
|
+
'Content-Type': 'text/event-stream',
|
|
43
|
+
'Cache-Control': 'no-cache',
|
|
44
|
+
Connection: 'keep-alive',
|
|
45
|
+
});
|
|
46
|
+
res.flushHeaders();
|
|
47
|
+
}
|
|
48
|
+
let prompt = args.prompt;
|
|
49
|
+
if (typeof prompt !== 'string') {
|
|
50
|
+
throw new Error('Prompt must be a string');
|
|
51
|
+
}
|
|
52
|
+
let stop = args.stop ? args.stop : undefined;
|
|
53
|
+
if (typeof stop === 'string') {
|
|
54
|
+
stop = [stop];
|
|
55
|
+
}
|
|
56
|
+
const completionReq = omitEmptyValues({
|
|
57
|
+
model: args.model,
|
|
58
|
+
prompt: args.prompt,
|
|
59
|
+
temperature: args.temperature ? args.temperature : undefined,
|
|
60
|
+
stream: args.stream ? Boolean(args.stream) : false,
|
|
61
|
+
maxTokens: args.max_tokens ? args.max_tokens : undefined,
|
|
62
|
+
seed: args.seed ? args.seed : undefined,
|
|
63
|
+
stop,
|
|
64
|
+
frequencyPenalty: args.frequency_penalty
|
|
65
|
+
? args.frequency_penalty
|
|
66
|
+
: undefined,
|
|
67
|
+
presencePenalty: args.presence_penalty
|
|
68
|
+
? args.presence_penalty
|
|
69
|
+
: undefined,
|
|
70
|
+
tokenBias: args.logit_bias ? args.logit_bias : undefined,
|
|
71
|
+
topP: args.top_p ? args.top_p : undefined,
|
|
72
|
+
// additional non-spec params
|
|
73
|
+
repeatPenaltyNum: args.repeat_penalty_num
|
|
74
|
+
? args.repeat_penalty_num
|
|
75
|
+
: undefined,
|
|
76
|
+
minP: args.min_p ? args.min_p : undefined,
|
|
77
|
+
topK: args.top_k ? args.top_k : undefined,
|
|
78
|
+
});
|
|
79
|
+
const { instance, release } = await modelServer.requestInstance(completionReq, controller.signal);
|
|
80
|
+
const task = instance.processTextCompletionTask(completionReq, {
|
|
81
|
+
signal: controller.signal,
|
|
82
|
+
onChunk: (chunk) => {
|
|
83
|
+
if (args.stream) {
|
|
84
|
+
const chunkData = {
|
|
85
|
+
id: task.id,
|
|
86
|
+
model: task.model,
|
|
87
|
+
object: 'text_completion',
|
|
88
|
+
created: Math.floor(task.createdAt.getTime() / 1000),
|
|
89
|
+
choices: [
|
|
90
|
+
{
|
|
91
|
+
index: 0,
|
|
92
|
+
text: chunk.text,
|
|
93
|
+
logprobs: null,
|
|
94
|
+
// @ts-ignore official api returns null here in the same case
|
|
95
|
+
finish_reason: null,
|
|
96
|
+
},
|
|
97
|
+
],
|
|
98
|
+
};
|
|
99
|
+
res.write(`data: ${JSON.stringify(chunkData)}\n\n`);
|
|
100
|
+
}
|
|
101
|
+
},
|
|
102
|
+
});
|
|
103
|
+
const result = await task.result;
|
|
104
|
+
release();
|
|
105
|
+
if (args.stream) {
|
|
106
|
+
if (args.stream_options?.include_usage) {
|
|
107
|
+
const finalChunk = {
|
|
108
|
+
id: task.id,
|
|
109
|
+
model: task.model,
|
|
110
|
+
object: 'text_completion',
|
|
111
|
+
created: Math.floor(task.createdAt.getTime() / 1000),
|
|
112
|
+
choices: [
|
|
113
|
+
{
|
|
114
|
+
index: 0,
|
|
115
|
+
text: '',
|
|
116
|
+
logprobs: null,
|
|
117
|
+
// @ts-ignore
|
|
118
|
+
finish_reason: result.finishReason
|
|
119
|
+
? finishReasonMap[result.finishReason]
|
|
120
|
+
: 'stop',
|
|
121
|
+
},
|
|
122
|
+
],
|
|
123
|
+
};
|
|
124
|
+
res.write(`data: ${JSON.stringify(finalChunk)}\n\n`);
|
|
125
|
+
}
|
|
126
|
+
res.write('data: [DONE]');
|
|
127
|
+
res.end();
|
|
128
|
+
}
|
|
129
|
+
else {
|
|
130
|
+
const response = {
|
|
131
|
+
id: task.id,
|
|
132
|
+
model: task.model,
|
|
133
|
+
object: 'text_completion',
|
|
134
|
+
created: Math.floor(task.createdAt.getTime() / 1000),
|
|
135
|
+
system_fingerprint: instance.fingerprint,
|
|
136
|
+
choices: [
|
|
137
|
+
{
|
|
138
|
+
index: 0,
|
|
139
|
+
text: result.text,
|
|
140
|
+
logprobs: null,
|
|
141
|
+
// @ts-ignore
|
|
142
|
+
finish_reason: result.finishReason
|
|
143
|
+
? finishReasonMap[result.finishReason]
|
|
144
|
+
: 'stop',
|
|
145
|
+
},
|
|
146
|
+
],
|
|
147
|
+
usage: {
|
|
148
|
+
prompt_tokens: result.promptTokens,
|
|
149
|
+
completion_tokens: result.completionTokens,
|
|
150
|
+
total_tokens: result.contextTokens,
|
|
151
|
+
},
|
|
152
|
+
};
|
|
153
|
+
res.writeHead(200, { 'Content-Type': 'application/json' });
|
|
154
|
+
res.end(JSON.stringify(response, null, 2));
|
|
155
|
+
}
|
|
156
|
+
}
|
|
157
|
+
catch (err) {
|
|
158
|
+
console.error(err);
|
|
159
|
+
if (args.stream) {
|
|
160
|
+
res.write('data: [ERROR]');
|
|
161
|
+
}
|
|
162
|
+
else {
|
|
163
|
+
res.writeHead(500, { 'Content-Type': 'application/json' });
|
|
164
|
+
res.end(JSON.stringify({ error: 'Internal server error' }));
|
|
165
|
+
}
|
|
166
|
+
}
|
|
167
|
+
};
|
|
168
|
+
}
|
|
169
|
+
//# sourceMappingURL=completions.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"completions.js","sourceRoot":"","sources":["../../../../src/api/openai/handlers/completions.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,oBAAoB,EAAE,MAAM,sCAAsC,CAAA;AAC3E,OAAO,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAA;AACtD,OAAO,EAAE,eAAe,EAAE,MAAM,aAAa,CAAA;AAc7C,iBAAiB;AACjB,oEAAoE;AACpE,MAAM,UAAU,uBAAuB,CAAC,WAAwB;IAC/D,OAAO,KAAK,EAAE,GAAoB,EAAE,GAAmB,EAAE,EAAE;QAC1D,IAAI,IAA4B,CAAA;QAEhC,IAAI,CAAC;YACJ,MAAM,IAAI,GAAG,MAAM,oBAAoB,CAAC,GAAG,CAAC,CAAA;YAC5C,IAAI,GAAG,IAAI,CAAA;QACZ,CAAC;QAAC,OAAO,CAAC,EAAE,CAAC;YACZ,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,CAAA;YAChB,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,iBAAiB,EAAE,CAAC,CAAC,CAAA;YACrD,OAAM;QACP,CAAC;QAED,8BAA8B;QAC9B,IAAI,CAAC,IAAI,CAAC,KAAK,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;YACjC,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,iBAAiB,EAAE,CAAC,CAAC,CAAA;YACrD,OAAM;QACP,CAAC;QACD,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;YAC1C,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,eAAe,EAAE,CAAC,CAAC,CAAA;YACnD,OAAM;QACP,CAAC;QAED,MAAM,UAAU,GAAG,IAAI,eAAe,EAAE,CAAA;QACxC,GAAG,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;YACpB,OAAO,CAAC,KAAK,CAAC,0BAA0B,CAAC,CAAA;YACzC,UAAU,CAAC,KAAK,EAAE,CAAA;QACnB,CAAC,CAAC,CAAA;QACF,GAAG,CAAC,EAAE,CAAC,KAAK,EAAE,GAAG,EAAE;YAClB,OAAO,CAAC,KAAK,CAAC,yBAAyB,CAAC,CAAA;YACxC,UAAU,CAAC,KAAK,EAAE,CAAA;QACnB,CAAC,CAAC,CAAA;QAEF,IAAI,CAAC;YACJ,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;gBACjB,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE;oBAClB,cAAc,EAAE,mBAAmB;oBACnC,eAAe,EAAE,UAAU;oBAC3B,UAAU,EAAE,YAAY;iBACxB,CAAC,CAAA;gBACF,GAAG,CAAC,YAAY,EAAE,CAAA;YACnB,CAAC;YAED,IAAI,MAAM,GAAG,IAAI,CAAC,MAAM,CAAA;YAExB,IAAI,OAAO,MAAM,KAAK,QAAQ,EAAE,CAAC;gBAChC,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAA;YAC3C,CAAC;YAED,IAAI,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS,CAAA;YAC5C,IAAI,OAAO,IAAI,KAAK,QAAQ,EAAE,CAAC;gBAC9B,IAAI,GAAG,CAAC,IAAI,CAAC,CAAA;YACd,CAAC;YAED,MAAM,aAAa,GAAG,eAAe,CAAwB;gBAC5D,KAAK,EAAE,IAAI,CAAC,KAAK;gBACjB,MAAM,EAAE,IAAI,CAAC,MAAgB;gBAC7B,WAAW,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC,SAAS;gBAC5D,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,KAAK;gBAClD,SAAS,EAAE,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,SAAS;gBACxD,IAAI,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS;gBACvC,IAAI;gBACJ,gBAAgB,EAAE,IAAI,CAAC,iBAAiB;oBACvC,CAAC,CAAC,IAAI,CAAC,iBAAiB;oBACxB,CAAC,CAAC,SAAS;gBACZ,eAAe,EAAE,IAAI,CAAC,gBAAgB;oBACrC,CAAC,CAAC,IAAI,CAAC,gBAAgB;oBACvB,CAAC,CAAC,SAAS;gBACZ,SAAS,EAAE,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,SAAS;gBACxD,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,SAAS;gBACzC,6BAA6B;gBAC7B,gBAAgB,EAAE,IAAI,CAAC,kBAAkB;oBACxC,CAAC,CAAC,IAAI,CAAC,kBAAkB;oBACzB,CAAC,CAAC,SAAS;gBACZ,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,SAAS;gBACzC,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,SAAS;aACzC,CAAC,CAAA;YAEF,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,MAAM,WAAW,CAAC,eAAe,CAC9D,aAAa,EACb,UAAU,CAAC,MAAM,CACjB,CAAA;YACD,MAAM,IAAI,GAAG,QAAQ,CAAC,yBAAyB,CAAC,aAAa,EAAE;gBAC9D,MAAM,EAAE,UAAU,CAAC,MAAM;gBACzB,OAAO,EAAE,CAAC,KAAK,EAAE,EAAE;oBAClB,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;wBACjB,MAAM,SAAS,GAA0B;4BACxC,EAAE,EAAE,IAAI,CAAC,EAAE;4BACX,KAAK,EAAE,IAAI,CAAC,KAAK;4BACjB,MAAM,EAAE,iBAAiB;4BACzB,OAAO,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,OAAO,EAAE,GAAG,IAAI,CAAC;4BACpD,OAAO,EAAE;gCACR;oCACC,KAAK,EAAE,CAAC;oCACR,IAAI,EAAE,KAAK,CAAC,IAAI;oCAChB,QAAQ,EAAE,IAAI;oCACd,6DAA6D;oCAC7D,aAAa,EAAE,IAAI;iCACnB;6BACD;yBACD,CAAA;wBACD,GAAG,CAAC,KAAK,CAAC,SAAS,IAAI,CAAC,SAAS,CAAC,SAAS,CAAC,MAAM,CAAC,CAAA;oBACpD,CAAC;gBACF,CAAC;aACD,CAAC,CAAA;YACF,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,MAAM,CAAA;YAChC,OAAO,EAAE,CAAA;YAET,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;gBACjB,IAAI,IAAI,CAAC,cAAc,EAAE,aAAa,EAAE,CAAC;oBACxC,MAAM,UAAU,GAA0B;wBACzC,EAAE,EAAE,IAAI,CAAC,EAAE;wBACX,KAAK,EAAE,IAAI,CAAC,KAAK;wBACjB,MAAM,EAAE,iBAAiB;wBACzB,OAAO,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,OAAO,EAAE,GAAG,IAAI,CAAC;wBACpD,OAAO,EAAE;4BACR;gCACC,KAAK,EAAE,CAAC;gCACR,IAAI,EAAE,EAAE;gCACR,QAAQ,EAAE,IAAI;gCACd,aAAa;gCACb,aAAa,EAAE,MAAM,CAAC,YAAY;oCACjC,CAAC,CAAC,eAAe,CAAC,MAAM,CAAC,YAAY,CAAC;oCACtC,CAAC,CAAC,MAAM;6BACT;yBACD;qBACD,CAAA;oBACD,GAAG,CAAC,KAAK,CACR,SAAS,IAAI,CAAC,SAAS,CAAC,UAAU,CAAC,MAAM,CACzC,CAAA;gBACF,CAAC;gBACD,GAAG,CAAC,KAAK,CAAC,cAAc,CAAC,CAAA;gBACzB,GAAG,CAAC,GAAG,EAAE,CAAA;YACV,CAAC;iBAAM,CAAC;gBACP,MAAM,QAAQ,GAAkC;oBAC/C,EAAE,EAAE,IAAI,CAAC,EAAE;oBACX,KAAK,EAAE,IAAI,CAAC,KAAK;oBACjB,MAAM,EAAE,iBAAiB;oBACzB,OAAO,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,OAAO,EAAE,GAAG,IAAI,CAAC;oBACpD,kBAAkB,EAAE,QAAQ,CAAC,WAAW;oBACxC,OAAO,EAAE;wBACR;4BACC,KAAK,EAAE,CAAC;4BACR,IAAI,EAAE,MAAM,CAAC,IAAI;4BACjB,QAAQ,EAAE,IAAI;4BACd,aAAa;4BACb,aAAa,EAAE,MAAM,CAAC,YAAY;gCACjC,CAAC,CAAC,eAAe,CAAC,MAAM,CAAC,YAAY,CAAC;gCACtC,CAAC,CAAC,MAAM;yBACT;qBACD;oBACD,KAAK,EAAE;wBACN,aAAa,EAAE,MAAM,CAAC,YAAY;wBAClC,iBAAiB,EAAE,MAAM,CAAC,gBAAgB;wBAC1C,YAAY,EAAE,MAAM,CAAC,aAAa;qBAClC;iBACD,CAAA;gBACD,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;gBAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAA;YAC3C,CAAC;QACF,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACd,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAA;YAClB,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC;gBACjB,GAAG,CAAC,KAAK,CAAC,eAAe,CAAC,CAAA;YAC3B,CAAC;iBAAM,CAAC;gBACP,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;gBAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,uBAAuB,EAAE,CAAC,CAAC,CAAA;YAC5D,CAAC;QACF,CAAC;IACF,CAAC,CAAA;AACF,CAAC"}
|
|
@@ -0,0 +1,74 @@
|
|
|
1
|
+
import { parseJSONRequestBody } from '../../../api/parseJSONRequestBody.js';
|
|
2
|
+
import { omitEmptyValues } from '../../../lib/util.js';
|
|
3
|
+
// v1/embeddings
|
|
4
|
+
// https://platform.openai.com/docs/api-reference/embeddings
|
|
5
|
+
export function createEmbeddingsHandler(modelServer) {
|
|
6
|
+
return async (req, res) => {
|
|
7
|
+
let args;
|
|
8
|
+
try {
|
|
9
|
+
const body = await parseJSONRequestBody(req);
|
|
10
|
+
args = body;
|
|
11
|
+
}
|
|
12
|
+
catch (e) {
|
|
13
|
+
console.error(e);
|
|
14
|
+
res.writeHead(400, { 'Content-Type': 'application/json' });
|
|
15
|
+
res.end(JSON.stringify({ error: 'Invalid request' }));
|
|
16
|
+
return;
|
|
17
|
+
}
|
|
18
|
+
// TODO ajv schema validation?
|
|
19
|
+
if (!args.model || !args.input) {
|
|
20
|
+
res.writeHead(400, { 'Content-Type': 'application/json' });
|
|
21
|
+
res.end(JSON.stringify({ error: 'Invalid request' }));
|
|
22
|
+
return;
|
|
23
|
+
}
|
|
24
|
+
if (!modelServer.modelExists(args.model)) {
|
|
25
|
+
res.writeHead(400, { 'Content-Type': 'application/json' });
|
|
26
|
+
res.end(JSON.stringify({ error: 'Invalid model' }));
|
|
27
|
+
return;
|
|
28
|
+
}
|
|
29
|
+
const controller = new AbortController();
|
|
30
|
+
req.on('close', () => {
|
|
31
|
+
console.debug('Client closed connection');
|
|
32
|
+
controller.abort();
|
|
33
|
+
});
|
|
34
|
+
req.on('end', () => {
|
|
35
|
+
console.debug('Client ended connection');
|
|
36
|
+
controller.abort();
|
|
37
|
+
});
|
|
38
|
+
try {
|
|
39
|
+
let input = args.input;
|
|
40
|
+
if (typeof input !== 'string') {
|
|
41
|
+
throw new Error('Input must be a string');
|
|
42
|
+
}
|
|
43
|
+
const embeddingsReq = omitEmptyValues({
|
|
44
|
+
model: args.model,
|
|
45
|
+
input: args.input,
|
|
46
|
+
});
|
|
47
|
+
const { instance, release } = await modelServer.requestInstance(embeddingsReq, controller.signal);
|
|
48
|
+
const task = instance.processEmbeddingTask(embeddingsReq);
|
|
49
|
+
const result = await task.result;
|
|
50
|
+
release();
|
|
51
|
+
const response = {
|
|
52
|
+
model: instance.modelId,
|
|
53
|
+
object: 'list',
|
|
54
|
+
data: result.embeddings.map((embedding, index) => ({
|
|
55
|
+
embedding: Array.from(embedding),
|
|
56
|
+
index,
|
|
57
|
+
object: 'embedding',
|
|
58
|
+
})),
|
|
59
|
+
usage: {
|
|
60
|
+
prompt_tokens: result.inputTokens,
|
|
61
|
+
total_tokens: result.inputTokens,
|
|
62
|
+
},
|
|
63
|
+
};
|
|
64
|
+
res.writeHead(200, { 'Content-Type': 'application/json' });
|
|
65
|
+
res.end(JSON.stringify(response, null, 2));
|
|
66
|
+
}
|
|
67
|
+
catch (err) {
|
|
68
|
+
console.error(err);
|
|
69
|
+
res.writeHead(500, { 'Content-Type': 'application/json' });
|
|
70
|
+
res.end(JSON.stringify({ error: 'Internal server error' }));
|
|
71
|
+
}
|
|
72
|
+
};
|
|
73
|
+
}
|
|
74
|
+
//# sourceMappingURL=embeddings.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"embeddings.js","sourceRoot":"","sources":["../../../../src/api/openai/handlers/embeddings.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,oBAAoB,EAAE,MAAM,sCAAsC,CAAA;AAC3E,OAAO,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAA;AAKtD,gBAAgB;AAChB,4DAA4D;AAC5D,MAAM,UAAU,uBAAuB,CAAC,WAAwB;IAC/D,OAAO,KAAK,EAAE,GAAoB,EAAE,GAAmB,EAAE,EAAE;QAC1D,IAAI,IAA4B,CAAA;QAEhC,IAAI,CAAC;YACJ,MAAM,IAAI,GAAG,MAAM,oBAAoB,CAAC,GAAG,CAAC,CAAA;YAC5C,IAAI,GAAG,IAAI,CAAA;QACZ,CAAC;QAAC,OAAO,CAAC,EAAE,CAAC;YACZ,OAAO,CAAC,KAAK,CAAC,CAAC,CAAC,CAAA;YAChB,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,iBAAiB,EAAE,CAAC,CAAC,CAAA;YACrD,OAAM;QACP,CAAC;QAED,8BAA8B;QAC9B,IAAI,CAAC,IAAI,CAAC,KAAK,IAAI,CAAC,IAAI,CAAC,KAAK,EAAE,CAAC;YAChC,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,iBAAiB,EAAE,CAAC,CAAC,CAAA;YACrD,OAAM;QACP,CAAC;QACD,IAAI,CAAC,WAAW,CAAC,WAAW,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;YAC1C,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,eAAe,EAAE,CAAC,CAAC,CAAA;YACnD,OAAM;QACP,CAAC;QAED,MAAM,UAAU,GAAG,IAAI,eAAe,EAAE,CAAA;QACxC,GAAG,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;YACpB,OAAO,CAAC,KAAK,CAAC,0BAA0B,CAAC,CAAA;YACzC,UAAU,CAAC,KAAK,EAAE,CAAA;QACnB,CAAC,CAAC,CAAA;QACF,GAAG,CAAC,EAAE,CAAC,KAAK,EAAE,GAAG,EAAE;YAClB,OAAO,CAAC,KAAK,CAAC,yBAAyB,CAAC,CAAA;YACxC,UAAU,CAAC,KAAK,EAAE,CAAA;QACnB,CAAC,CAAC,CAAA;QAEF,IAAI,CAAC;YAGJ,IAAI,KAAK,GAAG,IAAI,CAAC,KAAK,CAAA;YAEtB,IAAI,OAAO,KAAK,KAAK,QAAQ,EAAE,CAAC;gBAC/B,MAAM,IAAI,KAAK,CAAC,wBAAwB,CAAC,CAAA;YAC1C,CAAC;YAED,MAAM,aAAa,GAAG,eAAe,CAAmB;gBACvD,KAAK,EAAE,IAAI,CAAC,KAAK;gBACjB,KAAK,EAAE,IAAI,CAAC,KAAe;aAC3B,CAAC,CAAA;YAEF,MAAM,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,MAAM,WAAW,CAAC,eAAe,CAC9D,aAAa,EACb,UAAU,CAAC,MAAM,CACjB,CAAA;YACD,MAAM,IAAI,GAAG,QAAQ,CAAC,oBAAoB,CAAC,aAAa,CAAC,CAAA;YACzD,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,MAAM,CAAA;YAChC,OAAO,EAAE,CAAA;YAET,MAAM,QAAQ,GAAmC;gBAChD,KAAK,EAAE,QAAQ,CAAC,OAAO;gBACvB,MAAM,EAAE,MAAM;gBACd,IAAI,EAAE,MAAM,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,SAAS,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC;oBAClD,SAAS,EAAE,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC;oBAChC,KAAK;oBACL,MAAM,EAAE,WAAW;iBACnB,CAAC,CAAC;gBACH,KAAK,EAAE;oBACN,aAAa,EAAE,MAAM,CAAC,WAAW;oBACjC,YAAY,EAAE,MAAM,CAAC,WAAW;iBAChC;aACD,CAAA;YACD,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAA;QAE3C,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACd,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAA;YAClB,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;YAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,KAAK,EAAE,uBAAuB,EAAE,CAAC,CAAC,CAAA;QAC5D,CAAC;IACF,CAAC,CAAA;AACF,CAAC"}
|
|
File without changes
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"images.js","sourceRoot":"","sources":["../../../../src/api/openai/handlers/images.ts"],"names":[],"mappings":";AACA,sBAAsB;AACtB,+DAA+D"}
|
|
@@ -0,0 +1,23 @@
|
|
|
1
|
+
// https://platform.openai.com/docs/api-reference/models/list
|
|
2
|
+
export function createModelsHandler(modelServer) {
|
|
3
|
+
return async (req, res) => {
|
|
4
|
+
const models = modelServer.store.getStatus();
|
|
5
|
+
const data = Object.entries(models).map(([id, info]) => {
|
|
6
|
+
// const lastModDate = new Date(info.source.lastModified)
|
|
7
|
+
// const created = Math.floor(lastModDate.getTime() / 1000)
|
|
8
|
+
// const dirPath = path.dirname(info.source.file);
|
|
9
|
+
// const lastDir = path.basename(dirPath);
|
|
10
|
+
// const baseName = path.basename(info.source.file);
|
|
11
|
+
const owned_by = info.engine; // + ':' + path.join(lastDir, baseName);
|
|
12
|
+
return {
|
|
13
|
+
object: 'model',
|
|
14
|
+
id,
|
|
15
|
+
created: 0,
|
|
16
|
+
owned_by,
|
|
17
|
+
};
|
|
18
|
+
});
|
|
19
|
+
res.writeHead(200, { 'Content-Type': 'application/json' });
|
|
20
|
+
res.end(JSON.stringify({ object: 'list', data }, null, 2));
|
|
21
|
+
};
|
|
22
|
+
}
|
|
23
|
+
//# sourceMappingURL=models.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"models.js","sourceRoot":"","sources":["../../../../src/api/openai/handlers/models.ts"],"names":[],"mappings":"AAKA,6DAA6D;AAC7D,MAAM,UAAU,mBAAmB,CAAC,WAAwB;IAC3D,OAAO,KAAK,EAAE,GAAoB,EAAE,GAAmB,EAAE,EAAE;QAE1D,MAAM,MAAM,GAAG,WAAW,CAAC,KAAK,CAAC,SAAS,EAAE,CAAA;QAC5C,MAAM,IAAI,GAAmB,MAAM,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC,GAAG,CACtD,CAAC,CAAC,EAAE,EAAE,IAAI,CAAC,EAAE,EAAE;YACd,yDAAyD;YACzD,2DAA2D;YAE3D,kDAAkD;YAClD,0CAA0C;YAC1C,oDAAoD;YACpD,MAAM,QAAQ,GAAG,IAAI,CAAC,MAAM,CAAA,CAAA,wCAAwC;YAEpE,OAAO;gBACN,MAAM,EAAE,OAAO;gBACf,EAAE;gBACF,OAAO,EAAE,CAAC;gBACV,QAAQ;aACR,CAAA;QACF,CAAC,CACD,CAAA;QAED,GAAG,CAAC,SAAS,CAAC,GAAG,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE,CAAC,CAAA;QAC1D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAA;IAC3D,CAAC,CAAA;AACF,CAAC"}
|
|
File without changes
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"transcription.js","sourceRoot":"","sources":["../../../../src/api/openai/handlers/transcription.ts"],"names":[],"mappings":";AAAA,OAAO;AACP,2EAA2E"}
|
|
@@ -0,0 +1,7 @@
|
|
|
1
|
+
import type { ModelServer } from '../../server.js';
|
|
2
|
+
export declare function createOpenAIRequestHandlers(modelServer: ModelServer): {
|
|
3
|
+
chatCompletions: (req: import("http").IncomingMessage, res: import("http").ServerResponse) => Promise<void>;
|
|
4
|
+
completions: (req: import("http").IncomingMessage, res: import("http").ServerResponse) => Promise<void>;
|
|
5
|
+
models: (req: import("http").IncomingMessage, res: import("http").ServerResponse) => Promise<void>;
|
|
6
|
+
embeddings: (req: import("http").IncomingMessage, res: import("http").ServerResponse) => Promise<void>;
|
|
7
|
+
};
|
|
@@ -0,0 +1,14 @@
|
|
|
1
|
+
import { createChatCompletionHandler } from './handlers/chat.js';
|
|
2
|
+
import { createCompletionHandler } from './handlers/completions.js';
|
|
3
|
+
import { createModelsHandler } from './handlers/models.js';
|
|
4
|
+
import { createEmbeddingsHandler } from './handlers/embeddings.js';
|
|
5
|
+
// See OpenAI API specs at https://github.com/openai/openai-openapi/blob/master/openapi.yaml
|
|
6
|
+
export function createOpenAIRequestHandlers(modelServer) {
|
|
7
|
+
return {
|
|
8
|
+
chatCompletions: createChatCompletionHandler(modelServer),
|
|
9
|
+
completions: createCompletionHandler(modelServer),
|
|
10
|
+
models: createModelsHandler(modelServer),
|
|
11
|
+
embeddings: createEmbeddingsHandler(modelServer),
|
|
12
|
+
};
|
|
13
|
+
}
|
|
14
|
+
//# sourceMappingURL=index.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/api/openai/index.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,2BAA2B,EAAE,MAAM,oBAAoB,CAAA;AAChE,OAAO,EAAE,uBAAuB,EAAE,MAAM,2BAA2B,CAAA;AACnE,OAAO,EAAE,mBAAmB,EAAE,MAAM,sBAAsB,CAAA;AAC1D,OAAO,EAAE,uBAAuB,EAAE,MAAM,0BAA0B,CAAA;AAGlE,4FAA4F;AAC5F,MAAM,UAAU,2BAA2B,CAAC,WAAwB;IACnE,OAAO;QACN,eAAe,EAAE,2BAA2B,CAAC,WAAW,CAAC;QACzD,WAAW,EAAE,uBAAuB,CAAC,WAAW,CAAC;QACjD,MAAM,EAAE,mBAAmB,CAAC,WAAW,CAAC;QACxC,UAAU,EAAE,uBAAuB,CAAC,WAAW,CAAC;KAChD,CAAA;AACF,CAAC"}
|
|
@@ -0,0 +1,24 @@
|
|
|
1
|
+
export function parseJSONRequestBody(req) {
|
|
2
|
+
return new Promise((resolve, reject) => {
|
|
3
|
+
// if request is from express theres no need to parse anything
|
|
4
|
+
if ('body' in req) {
|
|
5
|
+
resolve(req.body);
|
|
6
|
+
return;
|
|
7
|
+
}
|
|
8
|
+
// for native http server
|
|
9
|
+
let body = '';
|
|
10
|
+
req.on('data', (chunk) => {
|
|
11
|
+
body += chunk.toString();
|
|
12
|
+
});
|
|
13
|
+
req.on('end', () => {
|
|
14
|
+
try {
|
|
15
|
+
const data = JSON.parse(body);
|
|
16
|
+
resolve(data);
|
|
17
|
+
}
|
|
18
|
+
catch (error) {
|
|
19
|
+
reject(error);
|
|
20
|
+
}
|
|
21
|
+
});
|
|
22
|
+
});
|
|
23
|
+
}
|
|
24
|
+
//# sourceMappingURL=parseJSONRequestBody.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"parseJSONRequestBody.js","sourceRoot":"","sources":["../../src/api/parseJSONRequestBody.ts"],"names":[],"mappings":"AAEA,MAAM,UAAU,oBAAoB,CAAC,GAA8B;IAClE,OAAO,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;QAEtC,8DAA8D;QAC9D,IAAI,MAAM,IAAI,GAAG,EAAE,CAAC;YACnB,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,CAAA;YACjB,OAAM;QACP,CAAC;QAED,yBAAyB;QACzB,IAAI,IAAI,GAAG,EAAE,CAAA;QACb,GAAG,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,KAAK,EAAE,EAAE;YACxB,IAAI,IAAI,KAAK,CAAC,QAAQ,EAAE,CAAA;QACzB,CAAC,CAAC,CAAA;QACF,GAAG,CAAC,EAAE,CAAC,KAAK,EAAE,GAAG,EAAE;YAClB,IAAI,CAAC;gBACJ,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;gBAC7B,OAAO,CAAC,IAAI,CAAC,CAAA;YACd,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBAChB,MAAM,CAAC,KAAK,CAAC,CAAA;YACd,CAAC;QACF,CAAC,CAAC,CAAA;IACH,CAAC,CAAC,CAAA;AACH,CAAC"}
|
|
@@ -0,0 +1,29 @@
|
|
|
1
|
+
import express from 'express';
|
|
2
|
+
export function createAPIMiddleware(server) {
|
|
3
|
+
const router = express.Router();
|
|
4
|
+
router.use((req, res, next) => {
|
|
5
|
+
console.debug('API call', {
|
|
6
|
+
method: req.method,
|
|
7
|
+
url: req.url,
|
|
8
|
+
body: req.body,
|
|
9
|
+
});
|
|
10
|
+
next();
|
|
11
|
+
});
|
|
12
|
+
return router;
|
|
13
|
+
// return async (req: IncomingMessage, res: ServerResponse) => {
|
|
14
|
+
// let args: any
|
|
15
|
+
// try {
|
|
16
|
+
// const body = await parseJSONRequestBody(req)
|
|
17
|
+
// args = body
|
|
18
|
+
// } catch (e) {
|
|
19
|
+
// console.error(e)
|
|
20
|
+
// res.writeHead(400, { 'Content-Type': 'application/json' })
|
|
21
|
+
// res.end(JSON.stringify({ error: 'Invalid request' }))
|
|
22
|
+
// return
|
|
23
|
+
// }
|
|
24
|
+
// console.debug('Handler', JSON.stringify(args, null, 2))
|
|
25
|
+
// res.writeHead(200, { 'Content-Type': 'application/json' })
|
|
26
|
+
// res.end(JSON.stringify({ message: 'Hello' }))
|
|
27
|
+
// }
|
|
28
|
+
}
|
|
29
|
+
//# sourceMappingURL=index.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/api/v1/index.ts"],"names":[],"mappings":"AACA,OAAO,OAAO,MAAM,SAAS,CAAA;AAI7B,MAAM,UAAU,mBAAmB,CAAC,MAAmB;IACtD,MAAM,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,CAAA;IAE/B,MAAM,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,EAAE;QAC7B,OAAO,CAAC,KAAK,CAAC,UAAU,EAAE;YACzB,MAAM,EAAE,GAAG,CAAC,MAAM;YAClB,GAAG,EAAE,GAAG,CAAC,GAAG;YACZ,IAAI,EAAE,GAAG,CAAC,IAAI;SACd,CAAC,CAAA;QACF,IAAI,EAAE,CAAA;IACP,CAAC,CAAC,CAAA;IACF,OAAO,MAAM,CAAA;IACb,gEAAgE;IAEhE,iBAAiB;IAEjB,SAAS;IACT,iDAAiD;IACjD,gBAAgB;IAChB,iBAAiB;IACjB,qBAAqB;IACrB,+DAA+D;IAC/D,0DAA0D;IAC1D,WAAW;IACX,KAAK;IAEL,2DAA2D;IAC3D,8DAA8D;IAC9D,iDAAiD;IAEjD,IAAI;AACL,CAAC"}
|
package/dist/cli.d.ts
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
#!/usr/bin/env node
|
package/dist/cli.js
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
1
|
+
#!/usr/bin/env node
|
|
2
|
+
"use strict";
|
|
3
|
+
process.stdout.write("Nothing here, yet.\n");
|
|
4
|
+
/* API Ideas:
|
|
5
|
+
- `infsrv ls` - list stored models
|
|
6
|
+
- `infsrv rm <name>` - remove a stored model
|
|
7
|
+
- `infsrv show <name>` - show details of a stored model
|
|
8
|
+
- `infsrv prepare <config>` - download everything required for given config
|
|
9
|
+
*/
|
|
10
|
+
//# sourceMappingURL=cli.js.map
|
package/dist/cli.js.map
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"cli.js","sourceRoot":"","sources":["../src/cli.ts"],"names":[],"mappings":";;AACA,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,sBAAsB,CAAC,CAAC;AAE7C;;;;;EAKE"}
|
|
@@ -0,0 +1,34 @@
|
|
|
1
|
+
import { InferenceModel, EmbeddingModel } from 'gpt4all';
|
|
2
|
+
import { EngineTextCompletionArgs, EngineChatCompletionArgs, EngineChatCompletionResult, EngineTextCompletionResult, EngineContext, EngineEmbeddingArgs, EngineEmbeddingResult, FileDownloadProgress, ModelConfig, TextCompletionParams, ChatMessage } from '../../types/index.js';
|
|
3
|
+
export type GPT4AllInstance = InferenceModel | EmbeddingModel;
|
|
4
|
+
export interface GPT4AllModelMeta {
|
|
5
|
+
url: string;
|
|
6
|
+
md5sum: string;
|
|
7
|
+
filename: string;
|
|
8
|
+
promptTemplate: string;
|
|
9
|
+
systemPrompt: string;
|
|
10
|
+
filesize: number;
|
|
11
|
+
ramrequired: number;
|
|
12
|
+
}
|
|
13
|
+
export interface GPT4AllModelConfig extends ModelConfig {
|
|
14
|
+
location: string;
|
|
15
|
+
md5?: string;
|
|
16
|
+
url?: string;
|
|
17
|
+
contextSize?: number;
|
|
18
|
+
batchSize?: number;
|
|
19
|
+
task: 'text-completion' | 'embedding';
|
|
20
|
+
initialMessages?: ChatMessage[];
|
|
21
|
+
completionDefaults?: TextCompletionParams;
|
|
22
|
+
device?: {
|
|
23
|
+
gpu?: boolean | 'auto' | (string & {});
|
|
24
|
+
gpuLayers?: number;
|
|
25
|
+
cpuThreads?: number;
|
|
26
|
+
};
|
|
27
|
+
}
|
|
28
|
+
export declare const autoGpu = true;
|
|
29
|
+
export declare function prepareModel({ config, log }: EngineContext<GPT4AllModelConfig>, onProgress?: (progress: FileDownloadProgress) => void, signal?: AbortSignal): Promise<GPT4AllModelMeta | undefined>;
|
|
30
|
+
export declare function createInstance({ config, log }: EngineContext<GPT4AllModelConfig>, signal?: AbortSignal): Promise<InferenceModel | EmbeddingModel>;
|
|
31
|
+
export declare function disposeInstance(instance: GPT4AllInstance): Promise<void>;
|
|
32
|
+
export declare function processTextCompletionTask({ request, config, onChunk }: EngineTextCompletionArgs<GPT4AllModelConfig>, instance: GPT4AllInstance, signal?: AbortSignal): Promise<EngineTextCompletionResult>;
|
|
33
|
+
export declare function processChatCompletionTask({ request, config, resetContext, log, onChunk }: EngineChatCompletionArgs<GPT4AllModelConfig>, instance: GPT4AllInstance, signal?: AbortSignal): Promise<EngineChatCompletionResult>;
|
|
34
|
+
export declare function processEmbeddingTask({ request, config }: EngineEmbeddingArgs, instance: GPT4AllInstance, signal?: AbortSignal): Promise<EngineEmbeddingResult>;
|