@retrivora-ai/rag-engine 2.4.3 → 2.4.6
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 +22 -0
- package/dist/{BatchProcessor-zbQMaZHm.d.mts → BatchProcessor-BGslIR5l.d.mts} +65 -4
- package/dist/{BatchProcessor-D42E706a.d.ts → BatchProcessor-C3gY6kZ0.d.ts} +65 -4
- package/dist/handlers/index.js +668 -837
- package/dist/handlers/index.mjs +678 -837
- package/dist/index.css +5 -0
- package/dist/index.d.mts +4 -28
- package/dist/index.d.ts +4 -28
- package/dist/index.js +91 -43
- package/dist/index.min.css +1 -1
- package/dist/index.mjs +91 -43
- package/dist/server.d.mts +21 -7
- package/dist/server.d.ts +21 -7
- package/dist/server.js +731 -875
- package/dist/server.mjs +741 -875
- package/package.json +11 -4
- package/src/LicensedRetrivora.ts +0 -278
- package/src/app/api/chat/route.ts +0 -4
- package/src/app/api/health/route.ts +0 -5
- package/src/app/api/ingest/route.ts +0 -27
- package/src/app/api/suggestions/route.ts +0 -5
- package/src/app/api/upload/route.ts +0 -5
- package/src/app/constants.tsx +0 -270
- package/src/app/favicon.ico +0 -0
- package/src/app/globals.css +0 -449
- package/src/app/layout.tsx +0 -22
- package/src/app/page.tsx +0 -64
- package/src/app/types.ts +0 -30
- package/src/browser.tsx +0 -127
- package/src/components/ChatWidget.tsx +0 -393
- package/src/components/ChatWindow.tsx +0 -952
- package/src/components/CodeViewer.tsx +0 -56
- package/src/components/ConfigProvider.tsx +0 -141
- package/src/components/DocumentUpload.tsx +0 -354
- package/src/components/DynamicChart.tsx +0 -297
- package/src/components/HourglassLoader.tsx +0 -87
- package/src/components/MarkdownComponents.tsx +0 -186
- package/src/components/MessageBubble.tsx +0 -498
- package/src/components/ObservabilityPanel.tsx +0 -374
- package/src/components/ProductCard.tsx +0 -107
- package/src/components/ProductCarousel.tsx +0 -97
- package/src/components/SourceCard.tsx +0 -61
- package/src/components/ThemeProvider.tsx +0 -8
- package/src/components/ThemeToggle.tsx +0 -29
- package/src/components/ThinkingBlock.tsx +0 -75
- package/src/components/UIDispatcher.tsx +0 -342
- package/src/components/VisualizationRenderer.tsx +0 -621
- package/src/components/constants.tsx +0 -503
- package/src/config/ConfigBuilder.ts +0 -285
- package/src/config/EmbeddingStrategy.ts +0 -148
- package/src/config/LicenseConfigResolver.ts +0 -298
- package/src/config/RagConfig.ts +0 -328
- package/src/config/UniversalProfiles.ts +0 -82
- package/src/config/constants.ts +0 -64
- package/src/config/serverConfig.ts +0 -378
- package/src/config/uiConstants.ts +0 -47
- package/src/core/BatchProcessor.ts +0 -385
- package/src/core/CircuitBreaker.ts +0 -118
- package/src/core/ConfigFetcher.ts +0 -220
- package/src/core/ConfigResolver.ts +0 -165
- package/src/core/ConfigValidator.ts +0 -197
- package/src/core/DatabaseStorage.ts +0 -526
- package/src/core/FreeTierLimitsGuard.ts +0 -281
- package/src/core/LLMRouter.ts +0 -79
- package/src/core/LangChainAgent.ts +0 -132
- package/src/core/LicenseValidator.ts +0 -284
- package/src/core/LicenseVerifier.ts +0 -261
- package/src/core/MultiAgentCoordinator.ts +0 -239
- package/src/core/Pipeline.ts +0 -1279
- package/src/core/PluginManager.ts +0 -277
- package/src/core/ProjectSettings.ts +0 -247
- package/src/core/ProviderHealthCheck.ts +0 -133
- package/src/core/ProviderInterfaces.ts +0 -37
- package/src/core/ProviderRegistry.ts +0 -156
- package/src/core/QueryProcessor.ts +0 -399
- package/src/core/Retrivora.ts +0 -98
- package/src/core/VectorPlugin.ts +0 -193
- package/src/core/mcp.ts +0 -261
- package/src/exceptions/index.ts +0 -131
- package/src/handlers/index.ts +0 -1625
- package/src/hooks/useHydrated.ts +0 -13
- package/src/hooks/useProjectSettings.ts +0 -105
- package/src/hooks/useRagChat.ts +0 -436
- package/src/hooks/useStoredMessages.ts +0 -64
- package/src/index.css +0 -4269
- package/src/index.ts +0 -137
- package/src/lib/plugin.ts +0 -24
- package/src/llm/ILLMProvider.ts +0 -60
- package/src/llm/LLMFactory.ts +0 -141
- package/src/llm/providers/AnthropicProvider.ts +0 -215
- package/src/llm/providers/GeminiProvider.ts +0 -311
- package/src/llm/providers/GroqProvider.ts +0 -184
- package/src/llm/providers/OllamaProvider.ts +0 -232
- package/src/llm/providers/OpenAIProvider.ts +0 -188
- package/src/llm/providers/QwenProvider.ts +0 -199
- package/src/llm/providers/UniversalLLMAdapter.ts +0 -422
- package/src/llm/utils.ts +0 -46
- package/src/providers/graphdb/BaseGraphProvider.ts +0 -43
- package/src/providers/graphdb/SimpleGraphProvider.ts +0 -66
- package/src/providers/vectordb/BaseVectorProvider.ts +0 -98
- package/src/providers/vectordb/ChromaDBProvider.ts +0 -130
- package/src/providers/vectordb/MilvusProvider.ts +0 -134
- package/src/providers/vectordb/MongoDBProvider.ts +0 -250
- package/src/providers/vectordb/MultiTablePostgresProvider.ts +0 -461
- package/src/providers/vectordb/PineconeProvider.ts +0 -295
- package/src/providers/vectordb/PostgreSQLProvider.ts +0 -227
- package/src/providers/vectordb/QdrantProvider.ts +0 -294
- package/src/providers/vectordb/RedisProvider.ts +0 -105
- package/src/providers/vectordb/UniversalVectorProvider.ts +0 -221
- package/src/providers/vectordb/WeaviateProvider.ts +0 -172
- package/src/rag/DocumentChunker.ts +0 -144
- package/src/rag/EntityExtractor.ts +0 -90
- package/src/rag/LlamaIndexIngestor.ts +0 -60
- package/src/rag/Reranker.ts +0 -106
- package/src/rendering/IntentClassifier.ts +0 -167
- package/src/rendering/RendererRegistry.ts +0 -164
- package/src/rendering/RuleEngine.ts +0 -67
- package/src/rendering/VisualizationDecisionEngine.ts +0 -164
- package/src/rendering/__tests__/VisualizationDecisionEngine.test.ts +0 -237
- package/src/rendering/index.ts +0 -12
- package/src/rendering/rules/Rule1SpecificInfoRule.ts +0 -28
- package/src/rendering/rules/Rule2ComparisonRule.ts +0 -24
- package/src/rendering/rules/Rule3ProductDiscoveryRule.ts +0 -28
- package/src/rendering/rules/Rule4AnalyticalRule.ts +0 -38
- package/src/rendering/rules/Rule5MixedResponseRule.ts +0 -29
- package/src/rendering/rules/Rule6SmallResultSetRule.ts +0 -24
- package/src/rendering/rules/Rule7LargeTableRule.ts +0 -35
- package/src/rendering/types.ts +0 -96
- package/src/server.ts +0 -156
- package/src/tailwind.css +0 -194
- package/src/telemetry/TelemetryService.ts +0 -330
- package/src/types/chat.ts +0 -89
- package/src/types/index.ts +0 -278
- package/src/types/pdf-parse.d.ts +0 -13
- package/src/types/props.ts +0 -169
- package/src/utils/DocumentParser.ts +0 -111
- package/src/utils/ProductExtractor.ts +0 -350
- package/src/utils/SchemaMapper.ts +0 -176
- package/src/utils/UITransformer.ts +0 -1812
- package/src/utils/synonyms.ts +0 -80
- package/src/utils/templateUtils.ts +0 -78
- package/src/version.ts +0 -6
|
@@ -1,232 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* Ollama LLM Provider (local / self-hosted)
|
|
3
|
-
*/
|
|
4
|
-
|
|
5
|
-
import axios, { AxiosInstance } from 'axios';
|
|
6
|
-
import { ILLMProvider, EmbedOptions } from '../ILLMProvider';
|
|
7
|
-
import { ChatMessage, ChatOptions } from '../../types';
|
|
8
|
-
import { LLMConfig, EmbeddingConfig } from '../../config/RagConfig';
|
|
9
|
-
import { IProviderValidator, IProviderHealthChecker, HealthCheckResult } from '../../core/ProviderInterfaces';
|
|
10
|
-
import { ValidationError } from '../../core/ConfigValidator';
|
|
11
|
-
import { buildSystemContent } from '../utils';
|
|
12
|
-
|
|
13
|
-
interface OllamaChatResponse {
|
|
14
|
-
message: { content: string };
|
|
15
|
-
}
|
|
16
|
-
|
|
17
|
-
interface OllamaEmbedResponse {
|
|
18
|
-
embedding: number[];
|
|
19
|
-
}
|
|
20
|
-
|
|
21
|
-
export class OllamaProvider implements ILLMProvider {
|
|
22
|
-
private readonly http: AxiosInstance;
|
|
23
|
-
private readonly llmConfig: LLMConfig;
|
|
24
|
-
private readonly embeddingConfig?: EmbeddingConfig;
|
|
25
|
-
|
|
26
|
-
constructor(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig) {
|
|
27
|
-
const rawBaseURL = (llmConfig.baseUrl ?? 'http://localhost:11434').replace(/\/+$/, '');
|
|
28
|
-
const baseURL = rawBaseURL.replace(/\/(?:api\/v1|v1|api)$/i, '');
|
|
29
|
-
const timeout = Number(llmConfig.options?.timeout) || 300_000;
|
|
30
|
-
const headers = (llmConfig.options?.headers as Record<string, string>) || {};
|
|
31
|
-
this.http = axios.create({ baseURL, timeout, headers });
|
|
32
|
-
this.llmConfig = llmConfig;
|
|
33
|
-
this.embeddingConfig = embeddingConfig;
|
|
34
|
-
}
|
|
35
|
-
|
|
36
|
-
static getValidator(): IProviderValidator {
|
|
37
|
-
return {
|
|
38
|
-
validate(config: Record<string, unknown>): ValidationError[] {
|
|
39
|
-
const errors: ValidationError[] = [];
|
|
40
|
-
if (!config.model) {
|
|
41
|
-
const isEmbedding = config.provider === 'ollama' && 'dimensions' in config;
|
|
42
|
-
const prefix = isEmbedding ? 'embedding' : 'llm';
|
|
43
|
-
errors.push({
|
|
44
|
-
field: `${prefix}.model`,
|
|
45
|
-
message: 'Ollama model name is required',
|
|
46
|
-
suggestion: 'e.g., "llama3" or "mistral"',
|
|
47
|
-
severity: 'error',
|
|
48
|
-
});
|
|
49
|
-
}
|
|
50
|
-
return errors;
|
|
51
|
-
}
|
|
52
|
-
};
|
|
53
|
-
}
|
|
54
|
-
|
|
55
|
-
static getHealthChecker(): IProviderHealthChecker {
|
|
56
|
-
return {
|
|
57
|
-
async check(config: Record<string, unknown>): Promise<HealthCheckResult> {
|
|
58
|
-
const timestamp = Date.now();
|
|
59
|
-
const baseUrl = (config.baseUrl as string) || 'http://localhost:11434';
|
|
60
|
-
const modelName = config.model as string;
|
|
61
|
-
|
|
62
|
-
try {
|
|
63
|
-
const axios = (await import('axios')).default;
|
|
64
|
-
const { data } = await axios.get(`${baseUrl}/api/tags`);
|
|
65
|
-
const models = (data.models as { name: string }[]) || [];
|
|
66
|
-
const hasModel = models.some((m) => m.name === modelName || m.name.startsWith(`${modelName}:`));
|
|
67
|
-
|
|
68
|
-
return {
|
|
69
|
-
healthy: true,
|
|
70
|
-
provider: 'ollama',
|
|
71
|
-
capabilities: {
|
|
72
|
-
baseUrl,
|
|
73
|
-
model: modelName,
|
|
74
|
-
available: hasModel,
|
|
75
|
-
totalModels: models.length
|
|
76
|
-
},
|
|
77
|
-
timestamp,
|
|
78
|
-
};
|
|
79
|
-
} catch {
|
|
80
|
-
return {
|
|
81
|
-
healthy: false,
|
|
82
|
-
provider: 'ollama',
|
|
83
|
-
error: `Ollama server not reachable at ${baseUrl}. Is it running?`,
|
|
84
|
-
timestamp,
|
|
85
|
-
};
|
|
86
|
-
}
|
|
87
|
-
}
|
|
88
|
-
};
|
|
89
|
-
}
|
|
90
|
-
|
|
91
|
-
async chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string> {
|
|
92
|
-
const basePrompt =
|
|
93
|
-
options?.systemPrompt ??
|
|
94
|
-
this.llmConfig.systemPrompt ??
|
|
95
|
-
'You are a helpful assistant. Use the provided context to answer the user\'s question.';
|
|
96
|
-
const system = buildSystemContent(basePrompt, context);
|
|
97
|
-
|
|
98
|
-
const { data } = await this.http.post<OllamaChatResponse>('/api/chat', {
|
|
99
|
-
model: this.llmConfig.model,
|
|
100
|
-
stream: false,
|
|
101
|
-
options: {
|
|
102
|
-
temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
|
|
103
|
-
num_predict: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
|
|
104
|
-
},
|
|
105
|
-
messages: [
|
|
106
|
-
{ role: 'system', content: system },
|
|
107
|
-
...messages.map((m) => ({ role: m.role, content: m.content })),
|
|
108
|
-
],
|
|
109
|
-
});
|
|
110
|
-
|
|
111
|
-
return data.message.content;
|
|
112
|
-
}
|
|
113
|
-
|
|
114
|
-
async *chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string> {
|
|
115
|
-
const basePrompt =
|
|
116
|
-
options?.systemPrompt ??
|
|
117
|
-
this.llmConfig.systemPrompt ??
|
|
118
|
-
'You are a helpful assistant. Use the provided context to answer the user\'s question.';
|
|
119
|
-
const system = buildSystemContent(basePrompt, context);
|
|
120
|
-
|
|
121
|
-
const response = await this.http.post('/api/chat', {
|
|
122
|
-
model: this.llmConfig.model,
|
|
123
|
-
stream: true,
|
|
124
|
-
options: {
|
|
125
|
-
temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
|
|
126
|
-
num_predict: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
|
|
127
|
-
},
|
|
128
|
-
messages: [
|
|
129
|
-
{ role: 'system', content: system },
|
|
130
|
-
...messages.map((m) => ({ role: m.role, content: m.content })),
|
|
131
|
-
],
|
|
132
|
-
}, { responseType: 'stream' });
|
|
133
|
-
|
|
134
|
-
let lineBuffer = '';
|
|
135
|
-
const stream = response.data;
|
|
136
|
-
if (!stream) {
|
|
137
|
-
throw new Error('[OllamaProvider] response.data is undefined for stream request. Axios might not support streams in this environment.');
|
|
138
|
-
}
|
|
139
|
-
|
|
140
|
-
for await (const chunk of stream) {
|
|
141
|
-
lineBuffer += chunk.toString();
|
|
142
|
-
const lines = lineBuffer.split('\n');
|
|
143
|
-
|
|
144
|
-
// Keep the last partial line in the buffer
|
|
145
|
-
lineBuffer = lines.pop() || '';
|
|
146
|
-
|
|
147
|
-
for (const line of lines) {
|
|
148
|
-
if (!line.trim()) continue;
|
|
149
|
-
try {
|
|
150
|
-
const json = JSON.parse(line);
|
|
151
|
-
if (json.message?.content) {
|
|
152
|
-
yield json.message.content;
|
|
153
|
-
}
|
|
154
|
-
if (json.done) return;
|
|
155
|
-
} catch {
|
|
156
|
-
// If parsing fails, it might be a malformed line or something else
|
|
157
|
-
// But with line buffering, this should be rare
|
|
158
|
-
console.warn('[OllamaProvider] Failed to parse streaming line:', line);
|
|
159
|
-
}
|
|
160
|
-
}
|
|
161
|
-
}
|
|
162
|
-
|
|
163
|
-
// Process anything left in the buffer at the end
|
|
164
|
-
if (lineBuffer.trim()) {
|
|
165
|
-
try {
|
|
166
|
-
const json = JSON.parse(lineBuffer);
|
|
167
|
-
if (json.message?.content) yield json.message.content;
|
|
168
|
-
} catch {
|
|
169
|
-
// Final cleanup
|
|
170
|
-
}
|
|
171
|
-
}
|
|
172
|
-
}
|
|
173
|
-
|
|
174
|
-
|
|
175
|
-
async embed(text: string, options?: EmbedOptions): Promise<number[]> {
|
|
176
|
-
const model =
|
|
177
|
-
options?.model ??
|
|
178
|
-
this.embeddingConfig?.model ??
|
|
179
|
-
'nomic-embed-text';
|
|
180
|
-
|
|
181
|
-
const baseURL =
|
|
182
|
-
this.embeddingConfig?.baseUrl ?? this.llmConfig.baseUrl ?? 'http://localhost:11434';
|
|
183
|
-
|
|
184
|
-
const client =
|
|
185
|
-
baseURL !== (this.llmConfig.baseUrl ?? 'http://localhost:11434')
|
|
186
|
-
? axios.create({ baseURL, timeout: 60_000 })
|
|
187
|
-
: this.http;
|
|
188
|
-
|
|
189
|
-
let prompt = text;
|
|
190
|
-
|
|
191
|
-
// Dynamically apply prefixes from configuration if they exist
|
|
192
|
-
// Fallback to standard nomic-embed-text prefixes if not provided
|
|
193
|
-
const queryPrefix = this.embeddingConfig?.queryPrefix ?? (model.includes('nomic') ? 'search_query: ' : '');
|
|
194
|
-
const docPrefix = this.embeddingConfig?.documentPrefix ?? (model.includes('nomic') ? 'search_document: ' : '');
|
|
195
|
-
|
|
196
|
-
if (options?.taskType === 'query' && queryPrefix) {
|
|
197
|
-
if (!prompt.startsWith(queryPrefix)) {
|
|
198
|
-
prompt = `${queryPrefix}${text}`;
|
|
199
|
-
}
|
|
200
|
-
} else if (options?.taskType === 'document' && docPrefix) {
|
|
201
|
-
if (!prompt.startsWith(docPrefix)) {
|
|
202
|
-
prompt = `${docPrefix}${text}`;
|
|
203
|
-
}
|
|
204
|
-
}
|
|
205
|
-
|
|
206
|
-
const { data } = await client.post<OllamaEmbedResponse>('/api/embeddings', {
|
|
207
|
-
model,
|
|
208
|
-
prompt,
|
|
209
|
-
});
|
|
210
|
-
|
|
211
|
-
return data.embedding;
|
|
212
|
-
}
|
|
213
|
-
|
|
214
|
-
async batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]> {
|
|
215
|
-
const vectors: number[][] = [];
|
|
216
|
-
// Sequential fallback for Ollama (batch support varies by version)
|
|
217
|
-
for (const text of texts) {
|
|
218
|
-
vectors.push(await this.embed(text, options));
|
|
219
|
-
}
|
|
220
|
-
return vectors;
|
|
221
|
-
}
|
|
222
|
-
|
|
223
|
-
async ping(): Promise<boolean> {
|
|
224
|
-
try {
|
|
225
|
-
await this.http.get('/api/tags');
|
|
226
|
-
return true;
|
|
227
|
-
} catch (err) {
|
|
228
|
-
console.error('[OllamaProvider] Ping failed:', err);
|
|
229
|
-
return false;
|
|
230
|
-
}
|
|
231
|
-
}
|
|
232
|
-
}
|
|
@@ -1,188 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* OpenAI LLM Provider
|
|
3
|
-
*/
|
|
4
|
-
|
|
5
|
-
import OpenAI from 'openai';
|
|
6
|
-
import { ILLMProvider, EmbedOptions } from '../ILLMProvider';
|
|
7
|
-
import { ChatMessage, ChatOptions } from '../../types';
|
|
8
|
-
import { LLMConfig, EmbeddingConfig } from '../../config/RagConfig';
|
|
9
|
-
import { IProviderValidator, IProviderHealthChecker, HealthCheckResult } from '../../core/ProviderInterfaces';
|
|
10
|
-
import { ValidationError } from '../../core/ConfigValidator';
|
|
11
|
-
import { buildSystemContent } from '../utils';
|
|
12
|
-
|
|
13
|
-
export class OpenAIProvider implements ILLMProvider {
|
|
14
|
-
private readonly client: OpenAI;
|
|
15
|
-
private readonly llmConfig: LLMConfig;
|
|
16
|
-
private readonly embeddingConfig?: EmbeddingConfig;
|
|
17
|
-
|
|
18
|
-
constructor(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig) {
|
|
19
|
-
if (!llmConfig.apiKey) throw new Error('[OpenAIProvider] llmConfig.apiKey is required');
|
|
20
|
-
this.client = new OpenAI({ apiKey: llmConfig.apiKey });
|
|
21
|
-
this.llmConfig = llmConfig;
|
|
22
|
-
this.embeddingConfig = embeddingConfig;
|
|
23
|
-
}
|
|
24
|
-
|
|
25
|
-
static getValidator(): IProviderValidator {
|
|
26
|
-
return {
|
|
27
|
-
validate(config: Record<string, unknown>): ValidationError[] {
|
|
28
|
-
const errors: ValidationError[] = [];
|
|
29
|
-
const isEmbedding = config.provider === 'openai' && 'dimensions' in config;
|
|
30
|
-
const prefix = isEmbedding ? 'embedding' : 'llm';
|
|
31
|
-
|
|
32
|
-
if (!config.apiKey) {
|
|
33
|
-
errors.push({
|
|
34
|
-
field: `${prefix}.apiKey`,
|
|
35
|
-
message: 'OpenAI API key is required',
|
|
36
|
-
suggestion: 'Set OPENAI_API_KEY environment variable',
|
|
37
|
-
severity: 'error',
|
|
38
|
-
});
|
|
39
|
-
}
|
|
40
|
-
if (!config.model) {
|
|
41
|
-
errors.push({
|
|
42
|
-
field: `${prefix}.model`,
|
|
43
|
-
message: 'OpenAI model name is required',
|
|
44
|
-
suggestion: isEmbedding ? 'e.g., "text-embedding-3-small"' : 'e.g., "gpt-4o"',
|
|
45
|
-
severity: 'error',
|
|
46
|
-
});
|
|
47
|
-
}
|
|
48
|
-
return errors;
|
|
49
|
-
}
|
|
50
|
-
};
|
|
51
|
-
}
|
|
52
|
-
|
|
53
|
-
static getHealthChecker(): IProviderHealthChecker {
|
|
54
|
-
return {
|
|
55
|
-
async check(config: Record<string, unknown>): Promise<HealthCheckResult> {
|
|
56
|
-
const timestamp = Date.now();
|
|
57
|
-
const apiKey = config.apiKey as string;
|
|
58
|
-
const modelName = config.model as string;
|
|
59
|
-
|
|
60
|
-
try {
|
|
61
|
-
const OpenAI = await import('openai');
|
|
62
|
-
const client = new OpenAI.default({ apiKey });
|
|
63
|
-
const models = await client.models.list();
|
|
64
|
-
const hasModel = models.data.some((m) => m.id === modelName);
|
|
65
|
-
return {
|
|
66
|
-
healthy: true,
|
|
67
|
-
provider: 'openai',
|
|
68
|
-
capabilities: { model: modelName, available: hasModel, totalModels: models.data.length },
|
|
69
|
-
timestamp,
|
|
70
|
-
};
|
|
71
|
-
} catch (error) {
|
|
72
|
-
return {
|
|
73
|
-
healthy: false,
|
|
74
|
-
provider: 'openai',
|
|
75
|
-
error: `Connection failed: ${error instanceof Error ? error.message : String(error)}`,
|
|
76
|
-
timestamp,
|
|
77
|
-
};
|
|
78
|
-
}
|
|
79
|
-
}
|
|
80
|
-
};
|
|
81
|
-
}
|
|
82
|
-
|
|
83
|
-
async chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string> {
|
|
84
|
-
const basePrompt =
|
|
85
|
-
options?.systemPrompt ??
|
|
86
|
-
this.llmConfig.systemPrompt ??
|
|
87
|
-
'You are a helpful assistant. Answer questions based on the provided context.';
|
|
88
|
-
|
|
89
|
-
const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
|
|
90
|
-
role: 'system',
|
|
91
|
-
content: buildSystemContent(basePrompt, context),
|
|
92
|
-
};
|
|
93
|
-
|
|
94
|
-
const safeMessages = (Array.isArray(messages) ? messages : [])
|
|
95
|
-
.filter((m): m is ChatMessage => Boolean(m && typeof m === 'object'))
|
|
96
|
-
.map((m) => ({
|
|
97
|
-
role: (m.role || 'user') as 'user' | 'assistant',
|
|
98
|
-
content: typeof m.content === 'string' ? m.content : (m.content != null ? String(m.content) : ''),
|
|
99
|
-
}));
|
|
100
|
-
|
|
101
|
-
const formattedMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
|
|
102
|
-
systemMessage,
|
|
103
|
-
...safeMessages,
|
|
104
|
-
];
|
|
105
|
-
|
|
106
|
-
const completion = await this.client.chat.completions.create({
|
|
107
|
-
model: this.llmConfig.model,
|
|
108
|
-
messages: formattedMessages,
|
|
109
|
-
max_tokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
|
|
110
|
-
temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
|
|
111
|
-
stop: options?.stop,
|
|
112
|
-
});
|
|
113
|
-
|
|
114
|
-
return completion.choices[0]?.message?.content ?? '';
|
|
115
|
-
}
|
|
116
|
-
|
|
117
|
-
async *chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string> {
|
|
118
|
-
const basePrompt =
|
|
119
|
-
options?.systemPrompt ??
|
|
120
|
-
this.llmConfig.systemPrompt ??
|
|
121
|
-
'You are a helpful assistant. Answer questions based on the provided context.';
|
|
122
|
-
|
|
123
|
-
const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
|
|
124
|
-
role: 'system',
|
|
125
|
-
content: buildSystemContent(basePrompt, context),
|
|
126
|
-
};
|
|
127
|
-
|
|
128
|
-
const safeStreamMessages = (Array.isArray(messages) ? messages : [])
|
|
129
|
-
.filter((m): m is ChatMessage => Boolean(m && typeof m === 'object'))
|
|
130
|
-
.map((m) => ({
|
|
131
|
-
role: (m.role || 'user') as 'user' | 'assistant',
|
|
132
|
-
content: typeof m.content === 'string' ? m.content : (m.content != null ? String(m.content) : ''),
|
|
133
|
-
}));
|
|
134
|
-
|
|
135
|
-
const formattedMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
|
|
136
|
-
systemMessage,
|
|
137
|
-
...safeStreamMessages,
|
|
138
|
-
];
|
|
139
|
-
|
|
140
|
-
const stream = await this.client.chat.completions.create({
|
|
141
|
-
model: this.llmConfig.model,
|
|
142
|
-
messages: formattedMessages,
|
|
143
|
-
max_tokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
|
|
144
|
-
temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
|
|
145
|
-
stop: options?.stop,
|
|
146
|
-
stream: true,
|
|
147
|
-
});
|
|
148
|
-
|
|
149
|
-
if (!stream) {
|
|
150
|
-
throw new Error('[OpenAIProvider] completions.create stream is undefined');
|
|
151
|
-
}
|
|
152
|
-
|
|
153
|
-
for await (const chunk of stream) {
|
|
154
|
-
const content = chunk.choices[0]?.delta?.content || '';
|
|
155
|
-
if (content) yield content;
|
|
156
|
-
}
|
|
157
|
-
}
|
|
158
|
-
|
|
159
|
-
async embed(text: string, options?: EmbedOptions): Promise<number[]> {
|
|
160
|
-
const results = await this.batchEmbed([text], options);
|
|
161
|
-
return results[0];
|
|
162
|
-
}
|
|
163
|
-
|
|
164
|
-
async batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]> {
|
|
165
|
-
const model =
|
|
166
|
-
options?.model ??
|
|
167
|
-
this.embeddingConfig?.model ??
|
|
168
|
-
'text-embedding-3-small';
|
|
169
|
-
|
|
170
|
-
const apiKey = this.embeddingConfig?.apiKey ?? this.llmConfig.apiKey;
|
|
171
|
-
const client = apiKey !== this.llmConfig.apiKey
|
|
172
|
-
? new OpenAI({ apiKey })
|
|
173
|
-
: this.client;
|
|
174
|
-
|
|
175
|
-
const response = await client.embeddings.create({ model, input: texts });
|
|
176
|
-
return response.data.map((d) => d.embedding);
|
|
177
|
-
}
|
|
178
|
-
|
|
179
|
-
async ping(): Promise<boolean> {
|
|
180
|
-
try {
|
|
181
|
-
await this.client.models.list();
|
|
182
|
-
return true;
|
|
183
|
-
} catch (err) {
|
|
184
|
-
console.error('[OpenAIProvider] Ping failed:', err);
|
|
185
|
-
return false;
|
|
186
|
-
}
|
|
187
|
-
}
|
|
188
|
-
}
|
|
@@ -1,199 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* Qwen LLM Provider (via Alibaba DashScope / ModelStudio compatible mode)
|
|
3
|
-
*/
|
|
4
|
-
|
|
5
|
-
import OpenAI from 'openai';
|
|
6
|
-
import { ILLMProvider, EmbedOptions } from '../ILLMProvider';
|
|
7
|
-
import { ChatMessage, ChatOptions } from '../../types';
|
|
8
|
-
import { LLMConfig, EmbeddingConfig } from '../../config/RagConfig';
|
|
9
|
-
import { IProviderValidator, IProviderHealthChecker, HealthCheckResult } from '../../core/ProviderInterfaces';
|
|
10
|
-
import { ValidationError } from '../../core/ConfigValidator';
|
|
11
|
-
import { buildSystemContent } from '../utils';
|
|
12
|
-
|
|
13
|
-
export class QwenProvider implements ILLMProvider {
|
|
14
|
-
private readonly client: OpenAI;
|
|
15
|
-
private readonly llmConfig: LLMConfig;
|
|
16
|
-
private readonly embeddingConfig?: EmbeddingConfig;
|
|
17
|
-
|
|
18
|
-
constructor(llmConfig: LLMConfig, embeddingConfig?: EmbeddingConfig) {
|
|
19
|
-
const apiKey = llmConfig.apiKey || process.env.DASHSCOPE_API_KEY || process.env.QWEN_API_KEY;
|
|
20
|
-
if (!apiKey) throw new Error('[QwenProvider] llmConfig.apiKey, DASHSCOPE_API_KEY, or QWEN_API_KEY environment variable is required');
|
|
21
|
-
|
|
22
|
-
this.client = new OpenAI({
|
|
23
|
-
apiKey,
|
|
24
|
-
baseURL: llmConfig.baseUrl || 'https://dashscope.aliyuncs.com/compatible-mode/v1',
|
|
25
|
-
});
|
|
26
|
-
this.llmConfig = llmConfig;
|
|
27
|
-
this.embeddingConfig = embeddingConfig;
|
|
28
|
-
}
|
|
29
|
-
|
|
30
|
-
static getValidator(): IProviderValidator {
|
|
31
|
-
return {
|
|
32
|
-
validate(config: Record<string, unknown>): ValidationError[] {
|
|
33
|
-
const errors: ValidationError[] = [];
|
|
34
|
-
const isEmbedding = config.provider === 'qwen' && 'dimensions' in config;
|
|
35
|
-
const prefix = isEmbedding ? 'embedding' : 'llm';
|
|
36
|
-
|
|
37
|
-
if (!config.apiKey && !process.env.DASHSCOPE_API_KEY && !process.env.QWEN_API_KEY) {
|
|
38
|
-
errors.push({
|
|
39
|
-
field: `${prefix}.apiKey`,
|
|
40
|
-
message: 'Qwen API key is required',
|
|
41
|
-
suggestion: 'Set DASHSCOPE_API_KEY or QWEN_API_KEY environment variable',
|
|
42
|
-
severity: 'error',
|
|
43
|
-
});
|
|
44
|
-
}
|
|
45
|
-
if (!config.model) {
|
|
46
|
-
errors.push({
|
|
47
|
-
field: `${prefix}.model`,
|
|
48
|
-
message: 'Qwen model name is required',
|
|
49
|
-
suggestion: isEmbedding ? 'e.g., "text-embedding-v3"' : 'e.g., "qwen-plus" or "qwen-max"',
|
|
50
|
-
severity: 'error',
|
|
51
|
-
});
|
|
52
|
-
}
|
|
53
|
-
return errors;
|
|
54
|
-
}
|
|
55
|
-
};
|
|
56
|
-
}
|
|
57
|
-
|
|
58
|
-
static getHealthChecker(): IProviderHealthChecker {
|
|
59
|
-
return {
|
|
60
|
-
async check(config: Record<string, unknown>): Promise<HealthCheckResult> {
|
|
61
|
-
const timestamp = Date.now();
|
|
62
|
-
const apiKey = (config.apiKey as string) || process.env.DASHSCOPE_API_KEY || process.env.QWEN_API_KEY || '';
|
|
63
|
-
const modelName = config.model as string;
|
|
64
|
-
|
|
65
|
-
try {
|
|
66
|
-
const OpenAI = await import('openai');
|
|
67
|
-
const client = new OpenAI.default({
|
|
68
|
-
apiKey,
|
|
69
|
-
baseURL: (config.baseUrl as string) || 'https://dashscope.aliyuncs.com/compatible-mode/v1',
|
|
70
|
-
});
|
|
71
|
-
const models = await client.models.list();
|
|
72
|
-
const hasModel = models.data.some((m) => m.id === modelName);
|
|
73
|
-
return {
|
|
74
|
-
healthy: true,
|
|
75
|
-
provider: 'qwen',
|
|
76
|
-
capabilities: { model: modelName, available: hasModel, totalModels: models.data.length },
|
|
77
|
-
timestamp,
|
|
78
|
-
};
|
|
79
|
-
} catch (error) {
|
|
80
|
-
return {
|
|
81
|
-
healthy: false,
|
|
82
|
-
provider: 'qwen',
|
|
83
|
-
error: `Connection failed: ${error instanceof Error ? error.message : String(error)}`,
|
|
84
|
-
timestamp,
|
|
85
|
-
};
|
|
86
|
-
}
|
|
87
|
-
}
|
|
88
|
-
};
|
|
89
|
-
}
|
|
90
|
-
|
|
91
|
-
async chat(messages: ChatMessage[], context: string, options?: ChatOptions): Promise<string> {
|
|
92
|
-
const basePrompt =
|
|
93
|
-
options?.systemPrompt ??
|
|
94
|
-
this.llmConfig.systemPrompt ??
|
|
95
|
-
'You are a helpful assistant. Answer questions based on the provided context.';
|
|
96
|
-
|
|
97
|
-
const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
|
|
98
|
-
role: 'system',
|
|
99
|
-
content: buildSystemContent(basePrompt, context),
|
|
100
|
-
};
|
|
101
|
-
|
|
102
|
-
const safeMessages = (Array.isArray(messages) ? messages : [])
|
|
103
|
-
.filter((m): m is ChatMessage => Boolean(m && typeof m === 'object'))
|
|
104
|
-
.map((m) => ({
|
|
105
|
-
role: (m.role || 'user') as 'user' | 'assistant',
|
|
106
|
-
content: typeof m.content === 'string' ? m.content : (m.content != null ? String(m.content) : ''),
|
|
107
|
-
}));
|
|
108
|
-
|
|
109
|
-
const formattedMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
|
|
110
|
-
systemMessage,
|
|
111
|
-
...safeMessages,
|
|
112
|
-
];
|
|
113
|
-
|
|
114
|
-
const completion = await this.client.chat.completions.create({
|
|
115
|
-
model: this.llmConfig.model,
|
|
116
|
-
messages: formattedMessages,
|
|
117
|
-
max_tokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
|
|
118
|
-
temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
|
|
119
|
-
stop: options?.stop,
|
|
120
|
-
});
|
|
121
|
-
|
|
122
|
-
return completion.choices[0]?.message?.content ?? '';
|
|
123
|
-
}
|
|
124
|
-
|
|
125
|
-
async *chatStream(messages: ChatMessage[], context: string, options?: ChatOptions): AsyncIterable<string> {
|
|
126
|
-
const basePrompt =
|
|
127
|
-
options?.systemPrompt ??
|
|
128
|
-
this.llmConfig.systemPrompt ??
|
|
129
|
-
'You are a helpful assistant. Answer questions based on the provided context.';
|
|
130
|
-
|
|
131
|
-
const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
|
|
132
|
-
role: 'system',
|
|
133
|
-
content: buildSystemContent(basePrompt, context),
|
|
134
|
-
};
|
|
135
|
-
|
|
136
|
-
const safeStreamMessages = (Array.isArray(messages) ? messages : [])
|
|
137
|
-
.filter((m): m is ChatMessage => Boolean(m && typeof m === 'object'))
|
|
138
|
-
.map((m) => ({
|
|
139
|
-
role: (m.role || 'user') as 'user' | 'assistant',
|
|
140
|
-
content: typeof m.content === 'string' ? m.content : (m.content != null ? String(m.content) : ''),
|
|
141
|
-
}));
|
|
142
|
-
|
|
143
|
-
const formattedMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
|
|
144
|
-
systemMessage,
|
|
145
|
-
...safeStreamMessages,
|
|
146
|
-
];
|
|
147
|
-
|
|
148
|
-
const stream = await this.client.chat.completions.create({
|
|
149
|
-
model: this.llmConfig.model,
|
|
150
|
-
messages: formattedMessages,
|
|
151
|
-
max_tokens: options?.maxTokens ?? this.llmConfig.maxTokens ?? 1024,
|
|
152
|
-
temperature: options?.temperature ?? this.llmConfig.temperature ?? 0.7,
|
|
153
|
-
stop: options?.stop,
|
|
154
|
-
stream: true,
|
|
155
|
-
});
|
|
156
|
-
|
|
157
|
-
if (!stream) {
|
|
158
|
-
throw new Error('[QwenProvider] completions.create stream is undefined');
|
|
159
|
-
}
|
|
160
|
-
|
|
161
|
-
for await (const chunk of stream) {
|
|
162
|
-
const content = chunk.choices[0]?.delta?.content || '';
|
|
163
|
-
if (content) yield content;
|
|
164
|
-
}
|
|
165
|
-
}
|
|
166
|
-
|
|
167
|
-
async embed(text: string, options?: EmbedOptions): Promise<number[]> {
|
|
168
|
-
const results = await this.batchEmbed([text], options);
|
|
169
|
-
return results[0];
|
|
170
|
-
}
|
|
171
|
-
|
|
172
|
-
async batchEmbed(texts: string[], options?: EmbedOptions): Promise<number[][]> {
|
|
173
|
-
const model =
|
|
174
|
-
options?.model ??
|
|
175
|
-
this.embeddingConfig?.model ??
|
|
176
|
-
'text-embedding-v1';
|
|
177
|
-
|
|
178
|
-
const apiKey = this.embeddingConfig?.apiKey ?? this.llmConfig.apiKey;
|
|
179
|
-
const client = apiKey !== this.llmConfig.apiKey
|
|
180
|
-
? new OpenAI({
|
|
181
|
-
apiKey,
|
|
182
|
-
baseURL: this.embeddingConfig?.baseUrl || 'https://dashscope.aliyuncs.com/compatible-mode/v1',
|
|
183
|
-
})
|
|
184
|
-
: this.client;
|
|
185
|
-
|
|
186
|
-
const response = await client.embeddings.create({ model, input: texts });
|
|
187
|
-
return response.data.map((d) => d.embedding);
|
|
188
|
-
}
|
|
189
|
-
|
|
190
|
-
async ping(): Promise<boolean> {
|
|
191
|
-
try {
|
|
192
|
-
await this.client.models.list();
|
|
193
|
-
return true;
|
|
194
|
-
} catch (err) {
|
|
195
|
-
console.error('[QwenProvider] Ping failed:', err);
|
|
196
|
-
return false;
|
|
197
|
-
}
|
|
198
|
-
}
|
|
199
|
-
}
|