openalmanac 0.2.58 → 0.3.1

package/dist/setup.js

@@ -524,7 +524,7 @@ function printResult(agent, loginResult, mcpChanged, toolCount) {
     w(row(`  ${WHITE_BOLD}Next steps${RST}`));
     w(empty);
     w(row(`  ${BLUE}1.${RST}  Type ${WHITE_BOLD}claude${RST} to start Claude Code`));
-    w(row(`  ${BLUE}2.${RST}  Say ${BLUE}"Let's explore <topic> using Almanac"${RST}`));
+    w(row(`  ${BLUE}2.${RST}  Say ${BLUE}"I want to contribute/explore the founders-inc wiki"${RST}`));
     w(empty);
     w(`  ${BLUE_DIM}\u2570${"─".repeat(innerW)}\u256f${RST}`);
     w("");

package/dist/tools/research.js

@@ -20,30 +20,79 @@ function coerceJson(schema) {
     }, schema);
 }
 export function registerResearchTools(server) {
+    // Discriminated union on `source` — each source's schema only surfaces its own
+    // options, so agents get typed autocomplete instead of an opaque options bag.
+    //
+    // Adding a new source (e.g. hackernews) = append another z.object variant with
+    // its own `source: z.literal(...)` and per-source fields. No existing schema
+    // changes.
+    const WebSearchInput = z.object({
+        source: z.literal("web").describe("Generic web search via Google/Serper. Use for general references, news, docs."),
+        query: z.string().min(1).describe("Search terms. Supports quoted phrases and site: operators."),
+        limit: z.number().int().min(1).max(20).default(10).describe("Max results (1-20, default 10)."),
+    });
+    const RedditSearchInput = z.object({
+        source: z.literal("reddit").describe("Search Reddit — use when the user wants community perspectives, subreddit consensus, lived experiences, or ranked-by-engagement content. Goes through a residential proxy so it sees past Reddit's anti-scraping."),
+        subreddit: z.string().optional().describe("Subreddit name without the 'r/' prefix (e.g. 'Harvard'). Omit to search across all of Reddit. Case-insensitive."),
+        query: z.string().optional().describe("Optional full-text search terms. Omit to return the subreddit's sorted listing (top posts of the year, etc.)."),
+        sort: z.enum(["top", "hot", "new", "rising", "controversial", "relevance", "comments"])
+            .default("top")
+            .describe("For listings: 'top'/'hot'/'new'/'rising'/'controversial'. For searches: 'relevance'/'top'/'new'/'comments'. Default 'top'."),
+        time_range: z.enum(["hour", "day", "week", "month", "year", "all"])
+            .default("year")
+            .describe("Time window for top/controversial listings and all searches. Default 'year'."),
+        limit: z.number().int().min(1).max(100).default(25).describe("Max posts to return (1-100, default 25)."),
+    });
     server.addTool({
         name: "search_web",
-        description: "Search the web for sources to cite in articles. Use this to find references before writing. Requires API key. Rate limit: 10/min.",
-        parameters: z.object({
-            query: z.string().describe("Search terms"),
-            limit: z.number().default(10).describe("Max results (1-20, default 10)"),
-        }),
-        async execute({ query, limit }) {
+        description: "Search the web or a specific community source (Reddit). Discriminated on `source`:\n\n" +
+            "- `source: \"web\"` — general web search via Google. Use for news, docs, scholarly references.\n" +
+            "- `source: \"reddit\"` — Reddit-aware search returning posts with score, flair, num_comments, permalink. " +
+            "Use when the user is asking about community perspectives, subreddit consensus, or 'what do people think about X'.\n\n" +
+            "Each source exposes its own parameters — follow the schema for the source you pick. " +
+            "Rate limit: 10/min. Requires API key.",
+        parameters: z.discriminatedUnion("source", [WebSearchInput, RedditSearchInput]),
+        async execute(input) {
+            if (input.source === "reddit") {
+                const params = {
+                    sort: input.sort,
+                    time_range: input.time_range,
+                    limit: input.limit,
+                };
+                if (input.subreddit)
+                    params.subreddit = input.subreddit;
+                if (input.query)
+                    params.query = input.query;
+                const resp = await request("GET", "/api/research/reddit/search", {
+                    auth: true,
+                    params,
+                });
+                return JSON.stringify(await resp.json(), null, 2);
+            }
             const resp = await request("GET", "/api/research/search", {
                 auth: true,
-                params: { query, limit },
+                params: { query: input.query, limit: input.limit },
             });
             return JSON.stringify(await resp.json(), null, 2);
         },
     });
     server.addTool({
         name: "read_webpage",
-        description: "Fetch a webpage and return its content as markdown. Use this to read sources found via search_web before citing them in articles. Supports web pages, PDFs, and YouTube videos. Requires API key. Rate limit: 5/min.",
+        description: "Fetch a URL and return its content as markdown. Routes automatically based on URL:\n" +
+            "- **Reddit threads** (reddit.com/r/{sub}/comments/{id}/...) — returns the post plus top-level threaded comments with scores and authors, via a residential proxy.\n" +
+            "- **YouTube videos** — returns title, description, transcript when available.\n" +
+            "- **PDFs** — extracts text.\n" +
+            "- **LinkedIn posts/profiles** — uses the LinkedIn scraper.\n" +
+            "- **Everything else** — generic web scrape with Firecrawl/Jina fallback.\n\n" +
+            "Requires API key. Rate limit: 5/min.",
         parameters: z.object({
-            url: z.string().describe("URL to read"),
+            url: z.string().url().describe("Full URL to read. For Reddit threads, use the canonical reddit.com/r/{sub}/comments/{id}/... form."),
             max_length: z
                 .number()
+                .int()
+                .min(1000)
                 .default(20000)
-                .describe("Max characters to return (default 20000). Use higher values for long-form sources."),
+                .describe("Max characters to return (default 20000). Increase for long-form sources; response is truncated with a note when it hits the cap."),
         }),
         async execute({ url, max_length }) {
             const resp = await request("GET", "/api/research/read", {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openalmanac",
-  "version": "0.2.58",
+  "version": "0.3.1",
   "description": "OpenAlmanac — pull, edit, and push articles to the open knowledge base",
   "type": "module",
   "bin": {