File size: 8,077 Bytes
10c8635 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 |
import type { Express } from "express";
import { createServer, type Server } from "http";
import {
GoogleGenerativeAI,
type ChatSession,
type GenerateContentResult,
} from "@google/generative-ai";
import { marked } from "marked";
import { setupEnvironment } from "./env";
const env = setupEnvironment();
const genAI = new GoogleGenerativeAI(env.GOOGLE_API_KEY);
const model = genAI.getGenerativeModel({
model: "gemini-2.0-flash-exp",
generationConfig: {
temperature: 0.9,
topP: 1,
topK: 1,
maxOutputTokens: 2048,
},
});
// Store chat sessions in memory
const chatSessions = new Map<string, ChatSession>();
// Format raw text into proper markdown
async function formatResponseToMarkdown(
text: string | Promise<string>
): Promise<string> {
// Ensure we have a string to work with
const resolvedText = await Promise.resolve(text);
// First, ensure consistent newlines
let processedText = resolvedText.replace(/\r\n/g, "\n");
// Process main sections (lines that start with word(s) followed by colon)
processedText = processedText.replace(
/^([A-Za-z][A-Za-z\s]+):(\s*)/gm,
"## $1$2"
);
// Process sub-sections (any remaining word(s) followed by colon within text)
processedText = processedText.replace(
/(?<=\n|^)([A-Za-z][A-Za-z\s]+):(?!\d)/gm,
"### $1"
);
// Process bullet points
processedText = processedText.replace(/^[β’ββ]\s*/gm, "* ");
// Split into paragraphs
const paragraphs = processedText.split("\n\n").filter(Boolean);
// Process each paragraph
const formatted = paragraphs
.map((p) => {
// If it's a header or list item, preserve it
if (p.startsWith("#") || p.startsWith("*") || p.startsWith("-")) {
return p;
}
// Add proper paragraph formatting
return `${p}\n`;
})
.join("\n\n");
// Configure marked options for better header rendering
marked.setOptions({
gfm: true,
breaks: true,
});
// Convert markdown to HTML using marked
return marked.parse(formatted);
}
interface WebSource {
uri: string;
title: string;
}
interface GroundingChunk {
web?: WebSource;
}
interface TextSegment {
startIndex: number;
endIndex: number;
text: string;
}
interface GroundingSupport {
segment: TextSegment;
groundingChunkIndices: number[];
confidenceScores: number[];
}
interface GroundingMetadata {
groundingChunks: GroundingChunk[];
groundingSupports: GroundingSupport[];
searchEntryPoint?: any;
webSearchQueries?: string[];
}
export function registerRoutes(app: Express): Server {
// Search endpoint - creates a new chat session
app.get("/api/search", async (req, res) => {
try {
const query = req.query.q as string;
if (!query) {
return res.status(400).json({
message: "Query parameter 'q' is required",
});
}
// Create a new chat session with search capability
const chat = model.startChat({
tools: [
{
// @ts-ignore - google_search is a valid tool but not typed in the SDK yet
google_search: {},
},
],
});
// Generate content with search tool
const result = await chat.sendMessage(query);
const response = await result.response;
console.log(
"Raw Google API Response:",
JSON.stringify(
{
text: response.text(),
candidates: response.candidates,
groundingMetadata: response.candidates?.[0]?.groundingMetadata,
},
null,
2
)
);
const text = response.text();
// Format the response text to proper markdown/HTML
const formattedText = await formatResponseToMarkdown(text);
// Extract sources from grounding metadata
const sourceMap = new Map<
string,
{ title: string; url: string; snippet: string }
>();
// Get grounding metadata from response
const metadata = response.candidates?.[0]?.groundingMetadata as any;
if (metadata) {
const chunks = metadata.groundingChunks || [];
const supports = metadata.groundingSupports || [];
chunks.forEach((chunk: any, index: number) => {
if (chunk.web?.uri && chunk.web?.title) {
const url = chunk.web.uri;
if (!sourceMap.has(url)) {
// Find snippets that reference this chunk
const snippets = supports
.filter((support: any) =>
support.groundingChunkIndices.includes(index)
)
.map((support: any) => support.segment.text)
.join(" ");
sourceMap.set(url, {
title: chunk.web.title,
url: url,
snippet: snippets || "",
});
}
}
});
}
const sources = Array.from(sourceMap.values());
// Generate a session ID and store the chat
const sessionId = Math.random().toString(36).substring(7);
chatSessions.set(sessionId, chat);
res.json({
sessionId,
summary: formattedText,
sources,
});
} catch (error: any) {
console.error("Search error:", error);
res.status(500).json({
message:
error.message || "An error occurred while processing your search",
});
}
});
// Follow-up endpoint - continues existing chat session
app.post("/api/follow-up", async (req, res) => {
try {
const { sessionId, query } = req.body;
if (!sessionId || !query) {
return res.status(400).json({
message: "Both sessionId and query are required",
});
}
const chat = chatSessions.get(sessionId);
if (!chat) {
return res.status(404).json({
message: "Chat session not found",
});
}
// Send follow-up message in existing chat
const result = await chat.sendMessage(query);
const response = await result.response;
console.log(
"Raw Google API Follow-up Response:",
JSON.stringify(
{
text: response.text(),
candidates: response.candidates,
groundingMetadata: response.candidates?.[0]?.groundingMetadata,
},
null,
2
)
);
const text = response.text();
// Format the response text to proper markdown/HTML
const formattedText = await formatResponseToMarkdown(text);
// Extract sources from grounding metadata
const sourceMap = new Map<
string,
{ title: string; url: string; snippet: string }
>();
// Get grounding metadata from response
const metadata = response.candidates?.[0]?.groundingMetadata as any;
if (metadata) {
const chunks = metadata.groundingChunks || [];
const supports = metadata.groundingSupports || [];
chunks.forEach((chunk: any, index: number) => {
if (chunk.web?.uri && chunk.web?.title) {
const url = chunk.web.uri;
if (!sourceMap.has(url)) {
// Find snippets that reference this chunk
const snippets = supports
.filter((support: any) =>
support.groundingChunkIndices.includes(index)
)
.map((support: any) => support.segment.text)
.join(" ");
sourceMap.set(url, {
title: chunk.web.title,
url: url,
snippet: snippets || "",
});
}
}
});
}
const sources = Array.from(sourceMap.values());
res.json({
summary: formattedText,
sources,
});
} catch (error: any) {
console.error("Follow-up error:", error);
res.status(500).json({
message:
error.message ||
"An error occurred while processing your follow-up question",
});
}
});
const httpServer = createServer(app);
return httpServer;
}
|