orangeslice 2.0.5 → 2.1.1

package/README.md

@@ -64,11 +64,21 @@ const [companies, searchPage, ai] = await Promise.all([
       }
    })
 ]);
+
+const startups = await services.crunchbase.search({
+   sql: `
+      SELECT name, website_url, linkedin_url
+      FROM public.crunchbase_scraper_lean
+      WHERE operating_status = 'Active'
+      LIMIT 10
+   `
+});
 ```
 
 ## Service map
 
 - `services.company.linkedin.search/enrich`
+- `services.crunchbase.search` (returns rows array directly)
 - `services.company.getEmployeesFromLinkedin` (database-only B2B path)
 - `services.person.linkedin.search/enrich`
 - `services.web.search/batchSearch`

package/dist/cli.js

@@ -90,19 +90,60 @@ function writeAgentsGuide(destDir) {
         return;
     const content = `# orangeslice Agent Guide
 
-Use the docs in this folder as the source of truth for all orangeslice operations.
+You are a coding agent using orangeslice services for B2B research and enrichment tasks.
 
-## Mandatory Read Order
-1. \`./services/index.md\` - service map
-2. Relevant service docs under \`./services/**\`
-3. \`./prospecting/index.md\` for company/people discovery workflows
+Use these docs as the source of truth. If there is any conflict between your prior knowledge and these docs, follow these docs.
 
-## Execution Rules
-- Prefer \`services.*\` APIs from orangeslice.
-- Run independent calls in parallel with \`Promise.all\`.
+## Core Behavior
+- Focus on completing the user's requested outcome with working code and clear next steps.
+- Prefer direct execution over long explanations.
+- Be concise, factual, and deterministic.
+- Ask a clarifying question only when a missing detail blocks progress.
+
+## Package Setup (Do Not Guess)
+- Import from the package name, not a local file path:
+  - \`import { services } from "orangeslice"\`
+  - \`import { configure, services } from "orangeslice"\` when setting API key programmatically
+- Do NOT use \`import { services } from "./orangeslice"\` unless the user explicitly has a local wrapper file at that path.
+- \`npx orangeslice\` is a setup/bootstrap command (docs sync, package install, auth). It does NOT execute user app scripts.
+
+## Runtime Requirements
+- If writing standalone scripts that use top-level \`await\`, use ESM:
+  - Set \`"type": "module"\` in \`package.json\`, or
+  - Use \`.mjs\` files.
+- If the project is CommonJS and cannot switch to ESM, avoid top-level \`await\` and wrap async code in an async function.
+
+## Mandatory Read Order (Before writing code)
+1. \`./services/index.md\` - service map and capabilities
+2. Relevant docs under \`./services/**\` for every service you plan to call
+3. \`./prospecting/index.md\` when doing discovery or lead generation tasks
+
+Do not call a service before reading its documentation.
+
+## Service Selection Rules
+- Prefer \`services.*\` APIs from orangeslice over ad hoc scraping or unstructured web calls.
 - For LinkedIn discovery, default to \`services.web.search\` unless it is a strict indexed lookup.
 - For scraping structured repeated elements, use \`services.browser.execute\`.
 - For broad scraping by URL, use \`services.scrape.website\`.
+- Use \`services.ai.generateObject\` for structured extraction/classification with a JSON schema.
+
+## Execution Rules
+- Parallelize independent async calls with \`Promise.all\`.
+- Avoid serial \`await\` inside loops when calls can run concurrently.
+- Keep code simple and composable; prefer small transformations over complex control flow.
+- Validate required inputs before expensive service calls.
+- Return structured, machine-usable output whenever possible.
+
+## Reliability and Safety
+- Do not invent service methods, params, or response shapes.
+- If a call fails, report the likely cause and provide a concrete retry/fallback path.
+- Do not expose secrets, API keys, or sensitive credentials in responses.
+- Do not claim an action succeeded unless the result confirms it.
+
+## Response Style
+- Briefly state what you are going to do, then do it.
+- Summarize outputs and include only relevant details.
+- When useful, provide a short "next actions" list.
 `;
     fs.writeFileSync(AGENTS_FILE, content, "utf8");
 }

package/dist/crunchbase.d.ts

@@ -0,0 +1,10 @@
+export interface CrunchbaseSearchParams {
+    sql: string;
+    userId?: string;
+}
+/**
+ * Search the Crunchbase lean table using SQL.
+ *
+ * Returns rows directly (no envelope).
+ */
+export declare function crunchbaseSearch<T = Record<string, unknown>>(params: CrunchbaseSearchParams): Promise<T[]>;

package/dist/crunchbase.js

@@ -0,0 +1,13 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.crunchbaseSearch = crunchbaseSearch;
+const api_1 = require("./api");
+/**
+ * Search the Crunchbase lean table using SQL.
+ *
+ * Returns rows directly (no envelope).
+ */
+async function crunchbaseSearch(params) {
+    const data = await (0, api_1.post)("/execute/crunchbase-sql", { sql: params.sql });
+    return data.rows ?? [];
+}

package/dist/index.d.ts

@@ -2,6 +2,8 @@ export { configure } from "./api";
 export type { OrangesliceConfig } from "./api";
 export { linkedinSearch } from "./b2b";
 export type { LinkedInSearchParams, LinkedInSearchResponse } from "./b2b";
+export { crunchbaseSearch } from "./crunchbase";
+export type { CrunchbaseSearchParams } from "./crunchbase";
 export { webSearch, webBatchSearch } from "./serp";
 export type { WebSearchQuery, WebSearchResult, WebSearchResponse, BatchWebSearchParams } from "./serp";
 export { generateObject } from "./generateObject";
@@ -21,12 +23,16 @@ export type { PersonLinkedinFindUrlParams, CompanyLinkedinFindUrlParams, PersonC
 import { runApifyActor } from "./apify";
 import { linkedinSearch } from "./b2b";
 import { browserExecute } from "./browser";
+import { crunchbaseSearch } from "./crunchbase";
 import { personLinkedinEnrich, personLinkedinFindUrl, personContactGet, companyLinkedinEnrich, companyLinkedinFindUrl, companyGetEmployeesFromLinkedin, geoParseAddress, builtWithLookupDomain, builtWithRelationships, builtWithSearchByTech } from "./expansion";
 import { scrapeWebsite } from "./firecrawl";
 import { generateObject } from "./generateObject";
 import { googleMapsScrape } from "./googleMaps";
 import { webBatchSearch, webSearch } from "./serp";
 export declare const services: {
+    crunchbase: {
+        search: typeof crunchbaseSearch;
+    };
     company: {
         linkedin: {
             findUrl: typeof companyLinkedinFindUrl;

package/dist/index.js

@@ -1,10 +1,12 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.services = exports.builtWithSearchByTech = exports.builtWithRelationships = exports.builtWithLookupDomain = exports.geoParseAddress = exports.companyGetEmployeesFromLinkedin = exports.companyLinkedinFindUrl = exports.companyLinkedinEnrich = exports.personContactGet = exports.personLinkedinFindUrl = exports.personLinkedinEnrich = exports.PREDICT_LEADS_OPERATION_IDS = exports.predictLeads = exports.executePredictLeads = exports.googleMapsScrape = exports.runApifyActor = exports.browserExecute = exports.scrapeWebsite = exports.generateObject = exports.webBatchSearch = exports.webSearch = exports.linkedinSearch = exports.configure = void 0;
+exports.services = exports.builtWithSearchByTech = exports.builtWithRelationships = exports.builtWithLookupDomain = exports.geoParseAddress = exports.companyGetEmployeesFromLinkedin = exports.companyLinkedinFindUrl = exports.companyLinkedinEnrich = exports.personContactGet = exports.personLinkedinFindUrl = exports.personLinkedinEnrich = exports.PREDICT_LEADS_OPERATION_IDS = exports.predictLeads = exports.executePredictLeads = exports.googleMapsScrape = exports.runApifyActor = exports.browserExecute = exports.scrapeWebsite = exports.generateObject = exports.webBatchSearch = exports.webSearch = exports.crunchbaseSearch = exports.linkedinSearch = exports.configure = void 0;
 var api_1 = require("./api");
 Object.defineProperty(exports, "configure", { enumerable: true, get: function () { return api_1.configure; } });
 var b2b_1 = require("./b2b");
 Object.defineProperty(exports, "linkedinSearch", { enumerable: true, get: function () { return b2b_1.linkedinSearch; } });
+var crunchbase_1 = require("./crunchbase");
+Object.defineProperty(exports, "crunchbaseSearch", { enumerable: true, get: function () { return crunchbase_1.crunchbaseSearch; } });
 var serp_1 = require("./serp");
 Object.defineProperty(exports, "webSearch", { enumerable: true, get: function () { return serp_1.webSearch; } });
 Object.defineProperty(exports, "webBatchSearch", { enumerable: true, get: function () { return serp_1.webBatchSearch; } });
@@ -36,6 +38,7 @@ Object.defineProperty(exports, "builtWithSearchByTech", { enumerable: true, get:
 const apify_2 = require("./apify");
 const b2b_2 = require("./b2b");
 const browser_2 = require("./browser");
+const crunchbase_2 = require("./crunchbase");
 const expansion_2 = require("./expansion");
 const firecrawl_2 = require("./firecrawl");
 const generateObject_2 = require("./generateObject");
@@ -43,6 +46,9 @@ const googleMaps_2 = require("./googleMaps");
 const predictLeads_2 = require("./predictLeads");
 const serp_2 = require("./serp");
 exports.services = {
+    crunchbase: {
+        search: crunchbase_2.crunchbaseSearch
+    },
     company: {
         linkedin: {
             findUrl: expansion_2.companyLinkedinFindUrl,

package/docs/integrations/gmail/index.md

@@ -9,4 +9,4 @@ Typed functions for Gmail actions powered by Orange Slice Google integrations.
 ## Email
 
 - `integrations.gmail.sendEmail(input)` - Send an email through the connected Gmail account
-- Heavy rate limit: `sendEmail` is capped at **20 calls/day** per connected Gmail account
+- Heavy rate limit: `sendEmail` is capped at **40 calls/day** per connected Gmail account

package/docs/integrations/gmail/sendEmail.md

@@ -2,7 +2,7 @@
 
 Send an email from the connected Gmail account.
 
-> Rate limit note for AI: `integrations.gmail.sendEmail(...)` is heavily rate-limited to **20 calls/day** per connected Gmail account. Use sparingly and batch/aggregate where possible.
+> Rate limit note for AI: `integrations.gmail.sendEmail(...)` is heavily rate-limited to **40 calls/day** per connected Gmail account. Use sparingly and batch/aggregate where possible.
 
 ```typescript
 // Basic email
@@ -25,18 +25,18 @@ await integrations.gmail.sendEmail({
 
 ## Input
 
-| Parameter          | Type        | Required | Description |
-| ------------------ | ----------- | -------- | ----------- |
-| `recipient_email`  | `string`    | No\*     | Primary `To` recipient |
-| `extra_recipients` | `string[]`  | No       | Additional `To` recipients |
-| `cc`               | `string[]`  | No       | CC recipients |
-| `bcc`              | `string[]`  | No       | BCC recipients |
-| `subject`          | `string`    | No\*     | Email subject |
-| `body`             | `string`    | No\*     | Email body (plain text or HTML) |
-| `is_html`          | `boolean`   | No       | Set to `true` when body is HTML |
-| `from_email`       | `string`    | No       | Optional verified send-as alias |
-| `attachment`       | `object`    | No       | Optional attachment payload |
-| `user_id`          | `string`    | No       | Gmail user id (`"me"` by default) |
+| Parameter          | Type       | Required | Description                       |
+| ------------------ | ---------- | -------- | --------------------------------- |
+| `recipient_email`  | `string`   | No\*     | Primary `To` recipient            |
+| `extra_recipients` | `string[]` | No       | Additional `To` recipients        |
+| `cc`               | `string[]` | No       | CC recipients                     |
+| `bcc`              | `string[]` | No       | BCC recipients                    |
+| `subject`          | `string`   | No\*     | Email subject                     |
+| `body`             | `string`   | No\*     | Email body (plain text or HTML)   |
+| `is_html`          | `boolean`  | No       | Set to `true` when body is HTML   |
+| `from_email`       | `string`   | No       | Optional verified send-as alias   |
+| `attachment`       | `object`   | No       | Optional attachment payload       |
+| `user_id`          | `string`   | No       | Gmail user id (`"me"` by default) |
 
 \*Gmail requires at least one recipient (`recipient_email`, `cc`, or `bcc`) and at least one of `subject` or `body`.
 

package/docs/prospecting/index.md

@@ -16,6 +16,7 @@ description: Strategies for searching or finding people and companies. This is a
 Run queries with built-in filters when the criteria is searchable:
 
 - **Web search (`services.web.search`)** — **Default for LinkedIn**. Use for keywords, niche queries, fuzzy matching, anything descriptive.
+- **Crunchbase (`services.crunchbase.search`)** — **Default for funding data**. Use for funding-stage, round type, amount, date windows, and investor-backed company discovery.
 - **LinkedIn B2B DB** — **Indexed lookups ONLY:** company by domain/slug/ID, employees at a known company (by company_id), basic funding (2-table join). Everything else = web search. See [QUICK_REF](./linkedin_data/QUICK_REF.md).
 - **Google Maps** — industry, location, ratings
 - **LinkedIn job search** — job filters, titles
@@ -83,11 +84,11 @@ When using qualification columns, think Circle & Star:
 | Source                   | Use When                                                    | Limitations                                              |
 | ------------------------ | ----------------------------------------------------------- | -------------------------------------------------------- |
 | **Web Search (Default)** | **Everything else** — keywords, niche, fuzzy, specific      | Requires verification columns for false positives.       |
+| **Crunchbase (Funding Default)** | Funding-focused prospecting: stage, round type, amount, recency, investors | Best for funding intelligence; use other sources for non-funding discovery criteria. |
 | **PredictLeads**         | Company intelligence, buying signals, and structured company events at scale | Coverage varies by company/market; use web search for very niche long-tail discovery. |
 | **Niche Directory Scrape** | Well-defined categories with existing lists (see below)   | Requires finding the right directory first.              |
 | **LinkedIn B2B DB**      | **Indexed lookups ONLY:** company by domain/slug/ID, employees at known company, basic 2-table funding. | **3s hard max. No keyword search, no LATERAL, no 3+ table joins.** Everything else = web search. |
 | **Google Maps**          | Local/SMB, physical locations, restaurants, retail          | Limited to businesses with physical presence.            |
-| **NPI Database**         | Healthcare providers                                        | Healthcare only. Free.                                   |
 | **Apify Actors**         | Platform-specific scraping (Instagram, TikTok, job boards)  | Per-platform setup. May break with platform changes.     |
 
 ### PredictLeads: When It Is Better Than Everything Else
@@ -105,6 +106,12 @@ Prefer other sources when:
 - You need local storefront/SMB discovery -> use Google Maps
 - You need fast indexed LinkedIn lookups by known IDs/domain/company -> use LinkedIn B2B DB
 
+### Funding Prospecting Standard: Use Crunchbase First
+
+For any request centered on funding data (for example: "Series A fintech companies", "companies that raised in the last 12 months", "recently funded startups"), use `services.crunchbase.search` as the **standard/default source**.
+
+Use LinkedIn B2B DB funding joins only when the user explicitly needs a LinkedIn-only workflow or a narrow lookup tied to existing LinkedIn records. Otherwise, Crunchbase should be the first choice for funding-oriented discovery.
+
 ### Niche Directory Scraping — For Well-Defined Categories
 
 When users ask for companies in a **specific, well-defined niche** (e.g., "fast food chains", "Fortune 500 companies", "Y Combinator startups"), the best approach is often to **find and scrape a curated directory or list**.
@@ -260,27 +267,28 @@ Don't overthink it — just create 2-3 views that match the columns you built. S
 
 ## Examples
 
-| User Request                                 | Approach         | Why                                                                 |
-| -------------------------------------------- | ---------------- | ------------------------------------------------------------------- |
-| "AI CRM companies"                           | Web search       | Keyword query → `"AI CRM" site:linkedin.com/company`                |
-| "Fintech startups"                           | Web search       | Fuzzy/descriptive → `"fintech" "startup" site:linkedin.com/company` |
-| "SDRs at Series A companies"                 | Web search       | Specific criteria → `"SDR" "Series A" site:linkedin.com/in`         |
-| "Companies using Kubernetes"                 | Web search       | Technology match → `"Kubernetes" site:linkedin.com/company`         |
-| "VPs who worked at Google"                   | Web search       | Fuzzy history match → `"VP" "Google" site:linkedin.com/in`          |
-| "1000 software engineers in Bay Area"        | B2B DB           | Simple title + location + high volume                               |
-| "All healthcare companies 100-500 employees" | B2B DB           | Industry + size + high volume                                       |
-| "Fast food chains that..."                   | Directory scrape | Scrape Wikipedia list → `browser.execute`                           |
-| "Restaurants in Austin"                      | Google Maps      | Local/SMB with physical presence                                    |
-| "Companies hiring SDRs"                      | LinkedIn Jobs    | Job search with title filter                                        |
-| "Warehouses implementing WMS"                | Circle + columns | Pull logistics companies → add "WMS Score" column                   |
-| "Companies that recently switched CRMs"      | Circle + columns | Pull SaaS companies → add "CRM Change Signals" column               |
+| User Request                                 | Approach         | Why                                                                         |
+| -------------------------------------------- | ---------------- | --------------------------------------------------------------------------- |
+| "AI CRM companies"                           | Web search       | Keyword query → `"AI CRM" site:linkedin.com/company`                        |
+| "Fintech startups"                           | Web search       | Fuzzy/descriptive → `"fintech" "startup" site:linkedin.com/company`         |
+| "SDRs at Series A companies"                 | Web search       | Specific criteria → `"SDR" "Series A" site:linkedin.com/in`                 |
+| "Series A/B companies raised last year"      | Crunchbase       | Funding-specific discovery is best handled via `services.crunchbase.search` |
+| "Companies using Kubernetes"                 | Web search       | Technology match → `"Kubernetes" site:linkedin.com/company`                 |
+| "VPs who worked at Google"                   | Web search       | Fuzzy history match → `"VP" "Google" site:linkedin.com/in`                  |
+| "1000 software engineers in Bay Area"        | B2B DB           | Simple title + location + high volume                                       |
+| "All healthcare companies 100-500 employees" | B2B DB           | Industry + size + high volume                                               |
+| "Fast food chains that..."                   | Directory scrape | Scrape Wikipedia list → `browser.execute`                                   |
+| "Restaurants in Austin"                      | Google Maps      | Local/SMB with physical presence                                            |
+| "Companies hiring SDRs"                      | LinkedIn Jobs    | Job search with title filter                                                |
+| "Warehouses implementing WMS"                | Circle + columns | Pull logistics companies → add "WMS Score" column                           |
+| "Companies that recently switched CRMs"      | Circle + columns | Pull SaaS companies → add "CRM Change Signals" column                       |
 
 ---
 
 ## Tools
 
 - **LinkedIn:** `services.company.linkedin.search({ sql: "SELECT ... FROM linkedin_company ..." })`, `services.person.linkedin.search({ sql: "SELECT ... FROM linkedin_profile ..." })` — **Lookup tool only, 3s max, 2-table joins max. Use web search for anything else.**
-- **Healthcare:** `healthcare.npi`
+- **Funding:** `services.crunchbase.search({ sql: "SELECT ... FROM ... WHERE ..." })` — **Default for funding search and screening.**
 - **Local/SMB:** `googleMaps.scrape`
 - **Web:** `web.search` + `browser.execute`
 - **Platforms:** `services.apify.runActor`

package/docs/services/apify/runActor.md

@@ -17,8 +17,9 @@ type runActor = (params: {
 
 ## Credits & Pricing
 
-**Credits: variable (custom). ~5 credits reserved per expected item. Settled at exact `usageTotalUsd × 500`.**
+**Credits: variable (custom). Reserved based on estimated items + compute. Settled at exact `usageTotalUsd × 500`.**
 
+- **Reservation:** Tiered on `datasetListParams.limit` — first 1000 items at 5 credits ($0.01), beyond 1000 at 1 credit ($0.002). Minimum 50 credits.
 - **Allowed pricing models:** `FREE`, `PRICE_PER_DATASET_ITEM`, `PAY_PER_EVENT`
 - **Blocked:** `FLAT_PRICE_PER_MONTH` (rental actors) — will throw an error
 - **Credits conversion:** $0.002 per credit (e.g., $0.01 = 5 credits)

package/docs/services/company/linkedin/search.md

@@ -390,6 +390,38 @@ WHERE lc.company_size = '51-200 employees'
 - **Use `lc` alias** for company tables
 - **Default to US**: `lc.country_code = 'US'`
 
+## Return Type
+
+`services.company.linkedin.search()` returns an object envelope:
+
+```typescript
+{
+   rows: (Record < string, unknown > []);
+   count: number;
+}
+```
+
+- `rows`: Result rows from your SQL query, with exactly the columns you selected.
+- `count`: Number of rows returned in `rows`.
+
+Example:
+
+```typescript
+const searchResult = await services.company.linkedin.search({
+   sql: `
+    SELECT
+      lc.company_name,
+      lc.domain,
+      'https://www.linkedin.com/company/' || lc.universal_name AS lc_linkedin_url
+    FROM linkedin_company lc
+    WHERE lc.domain = 'stripe.com'
+    LIMIT 1
+  `
+});
+
+return searchResult.rows; // Most spreadsheet snippets should return rows
+```
+
 ---
 
 ## Table Aliases