feat: add state-level HTML/PDF adapters and updated capacity filters

.gitignore

@@ -1,3 +1,10 @@
+
+venv/
+*.pyc
+__pycache__/
+.DS_Store
+Makefile.old
+start-up-project.txt
 .venv/
 __pycache__/
 *.pyc

app/ingest.py

@@ -2,7 +2,7 @@
 from __future__ import annotations
 import json
 from pathlib import Path
-from typing import Dict, List, Any
+from typing import Dict, List, Any, Tuple, Optional
 
 import yaml
 import numpy as np
@@ -11,6 +11,12 @@ from sentence_transformers import SentenceTransformer
 from app.paths import DOCSTORE_DIR, INDEX_DIR
 from .normalize import normalize  # ← central normalizer
 
+import re
+import time
+import hashlib
+import requests
+from bs4 import BeautifulSoup
+
 
 # -------------------- Config --------------------
 
@@ -19,6 +25,65 @@ def load_config(cfg_path: str) -> Dict:
         return yaml.safe_load(f)
 
 
+# -------------------- Capacity / Geo Filters (config-driven) --------------------
+# controls live in config/sources.yaml:
+# filters:
+#   capacity_only: true
+#   pa_md_only: false
+
+_INCLUDE_PATTERNS = [re.compile(p, re.I) for p in [
+    r"\bcapacity(?:[-\s]?building)?\b",
+    r"\btechnical\s+assistance\b",
+    r"\bTA\b",
+    r"\borganizational\s+(capacity|effectiveness|development|readiness|stabilization)\b",
+    r"\borganization(?:al)?\s+infrastructure\b",
+    r"\bback[-\s]?office\b|\bbackbone\s+organization\b",
+    r"\bgovernance\b|\bboard\s+development\b|\bboard\s+training\b",
+    r"\bpre[-\s]?development\b|\bpredevelopment\b|\bplanning\s+grant\b",
+    r"\bdata\s+systems?\b|\bCRM\b|\bcase\s+management\b",
+    r"\b(staff|workforce)\s+capacity\b|\bhire\s+(?:staff|positions?)\b",
+    r"\bscal(?:e|ing)\s+capacity\b|\bexpand\s+capacity\b",
+    r"\bnonprofit\b|\bfaith[-\s]?based\b|\bcommunity[-\s]?based\b",
+]]
+
+_EXCLUDE_PATTERNS = [re.compile(p, re.I) for p in [
+    r"\bteaching\s+assistant\b|\bTAs\b",
+    r"\bbench\s+capacity\b|\bmanufacturing\s+capacity\b(?!.*organiz)",
+    r"\bclinical\s+trial\b|\blaboratory\s+capacity\b(?!.*community)",
+    r"\b(postsecondary|university|college)\b(?!.*community\s+partner)",
+    r"\bconstruction\b(?!.*(admin|organiz|back[-\s]?office|governance|systems))",
+]]
+
+_PA_MD_HINTS = re.compile(
+    r"\b("
+    r"Pennsylvania|PA\b|Harrisburg|Philadelphia|Allegheny|Montgomery County\b|Pittsburgh|Scranton|Erie|"
+    r"Maryland|MD\b|Annapolis|Baltimore|Prince\s+George'?s|Howard County\b"
+    r")\b",
+    re.I,
+)
+
+def _doc_text_from_row(rec: Dict[str, Any]) -> str:
+    title = rec.get("title") or ""
+    synopsis = rec.get("synopsis") or rec.get("summary") or ""
+    agency = rec.get("agency") or ""
+    eligibility = rec.get("eligibility") or ""
+    categories = " ".join(rec.get("categories") or []) if isinstance(rec.get("categories"), list) else (rec.get("categories") or "")
+    geo = rec.get("geo") or ""
+    return "\n".join([title, synopsis, agency, eligibility, categories, geo]).strip()
+
+def _is_capacity_building_text(text: str) -> bool:
+    if not text:
+        return False
+    if any(p.search(text) for p in _EXCLUDE_PATTERNS):
+        return False
+    return any(p.search(text) for p in _INCLUDE_PATTERNS)
+
+def _is_pa_md_text(text: str) -> bool:
+    if not text:
+        return False
+    return bool(_PA_MD_HINTS.search(text))
+
+
 # -------------------- Grants.gov collector --------------------
 
 def _collect_from_grantsgov_api(src: Dict) -> List[Dict[str, Any]]:
@@ -39,6 +104,194 @@ def _collect_from_grantsgov_api(src: Dict) -> List[Dict[str, Any]]:
     return [h for h in hits if isinstance(h, dict)]
 
 
+# -------------------- NEW: Generic HTML / PDF collectors --------------------
+
+_HTTP_HEADERS = {
+    "User-Agent": "grants-rag/1.0 (+https://example.local) requests",
+    "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
+}
+
+def _http_get(url: str, timeout: int = 20) -> Optional[requests.Response]:
+    try:
+        r = requests.get(url, headers=_HTTP_HEADERS, timeout=timeout)
+        if r.status_code == 200 and r.content:
+            return r
+    except requests.RequestException:
+        return None
+    return None
+
+def _soup(html: str) -> BeautifulSoup:
+    # use lxml or html5lib if available for robustness
+    return BeautifulSoup(html, "lxml")
+
+def _text_from_soup(s: BeautifulSoup, selectors: Optional[List[str]] = None) -> Tuple[str, str]:
+    """
+    Returns (title, text). Uses selectors if provided;
+    falls back to common content containers.
+    """
+    title = s.title.string.strip() if s.title and s.title.string else ""
+
+    nodes = []
+    if selectors:
+        for css in selectors:
+            nodes.extend(s.select(css) or [])
+    if not nodes:
+        for css in ("main", "article", "#content", ".content", "[role='main']"):
+            nodes.extend(s.select(css) or [])
+    if not nodes:
+        nodes = [s.body] if s.body else []
+
+    parts: List[str] = []
+    for n in nodes:
+        if not n:
+            continue
+        txt = n.get_text(separator="\n", strip=True)
+        if txt:
+            parts.append(txt)
+    body = "\n\n".join(parts).strip()
+    return title, body
+
+def _make_id(*fields: str) -> str:
+    h = hashlib.sha1()
+    for f in fields:
+        if f:
+            h.update(f.encode("utf-8", "ignore"))
+            h.update(b"|")
+    return h.hexdigest()
+
+def _normalize_web_record(
+    source_name: str,
+    url: str,
+    title: str,
+    body: str,
+    static: Dict[str, Any],
+    extra: Optional[Dict[str, Any]] = None,
+) -> Dict[str, Any]:
+    """
+    Produce a record shaped like normalize() output so downstream stays unchanged.
+    """
+    rec = {
+        "id": (extra or {}).get("id") or _make_id(url, title or body[:160]),
+        "title": title or (extra.get("title") if extra else "") or url,
+        "synopsis": body[:2000],  # clip; embeddings use title+synopsis later
+        "summary": None,
+        "url": url,
+        "source": source_name,
+        "geo": static.get("geo"),
+        "categories": static.get("categories"),
+        "agency": (extra or {}).get("agency", ""),
+        "eligibility": (extra or {}).get("eligibility", ""),
+        "deadline": (extra or {}).get("deadline"),
+        "program_number": (extra or {}).get("program_number"),
+        "posted_date": (extra or {}).get("posted_date"),
+    }
+    return rec
+
+def _collect_from_http_html(entry: Dict, source_name: str, static: Dict[str, Any]) -> List[Dict[str, Any]]:
+    """
+    Supports types: 'web_page' and 'http_html'
+    Config keys supported:
+      - url (str)
+      - parse: { follow_links: bool, link_selectors: [..], content_selectors: [..] }
+      - crawl: { schedule: "...", max_depth: int }  # max_depth 0/None = only landing
+    """
+    url = entry.get("url")
+    if not url:
+        return []
+    r = _http_get(url)
+    if not r:
+        return []
+
+    s = _soup(r.text)
+    parse = entry.get("parse", {}) or entry.get("extract", {}) or {}
+    content_selectors = parse.get("content_selectors") or []
+    title, body = _text_from_soup(s, content_selectors)
+
+    rows = []
+    rows.append(_normalize_web_record(source_name, url, title, body, static, extra={"posted_date": None}))
+
+    # follow links?
+    follow = bool(parse.get("follow_links"))
+    link_selectors = parse.get("link_selectors") or []
+    crawl = entry.get("crawl", {}) or {}
+    max_depth = int(crawl.get("max_depth", 0) or 0)
+    visited = set([url])
+
+    def _enq_links(soup: BeautifulSoup) -> List[str]:
+        if link_selectors:
+            links = []
+            for sel in link_selectors:
+                for a in soup.select(sel) or []:
+                    href = a.get("href")
+                    if href and href.startswith("http"):
+                        links.append(href)
+            out, seen = [], set()
+            for h in links:
+                if h not in seen:
+                    out.append(h)
+                    seen.add(h)
+            return out[:40]  # polite cap
+        return []
+
+    if follow and max_depth > 0:
+        frontier = _enq_links(s)
+        depth = 1
+        while frontier and depth <= max_depth and len(rows) < 200:
+            next_frontier = []
+            for link in frontier:
+                if link in visited:
+                    continue
+                visited.add(link)
+                rr = _http_get(link)
+                if not rr:
+                    continue
+                ss = _soup(rr.text)
+                t2, b2 = _text_from_soup(ss, content_selectors)
+                if b2:
+                    rows.append(_normalize_web_record(source_name, link, t2, b2, static, extra={"posted_date": None}))
+                if depth < max_depth:
+                    next_frontier.extend(_enq_links(ss))
+                time.sleep(0.1)  # gentle
+            frontier = next_frontier
+            depth += 1
+
+    return rows
+
+def _collect_from_http_pdf(entry: Dict, source_name: str, static: Dict[str, Any]) -> List[Dict[str, Any]]:
+    """
+    type: 'http_pdf'
+    keys:
+      - url  (single PDF fetch)
+    """
+    url = entry.get("url")
+    if not url:
+        return []
+
+    try:
+        from pdfminer.high_level import extract_text  # lazy import
+    except Exception:
+        return []
+
+    rows = []
+    r = _http_get(url, timeout=40)
+    if not r:
+        return rows
+    tmp = DOCSTORE_DIR / (hashlib.sha1(url.encode("utf-8")).hexdigest() + ".pdf")
+    try:
+        DOCSTORE_DIR.mkdir(parents=True, exist_ok=True)
+        tmp.write_bytes(r.content)
+        body = extract_text(str(tmp)) or ""
+    finally:
+        try:
+            tmp.unlink(missing_ok=True)
+        except Exception:
+            pass
+    title = entry.get("name") or "PDF Document"
+    if body.strip():
+        rows.append(_normalize_web_record(source_name, url, title, body, static, extra={"posted_date": None}))
+    return rows
+
+
 # -------------------- Write docstore & build index --------------------
 
 def _save_docstore(recs: List[Dict[str, Any]]) -> str:
@@ -55,7 +308,6 @@ def _build_index_from_docstore() -> int:
     if not ds_path.exists():
         raise RuntimeError("Docstore not found. Run ingest first.")
 
-    # Load records → texts + metas
     texts: List[str] = []
     metas: List[Dict[str, Any]] = []
     with ds_path.open("r", encoding="utf-8") as f:
@@ -85,16 +337,15 @@ def _build_index_from_docstore() -> int:
     print(f"[index] Rows loaded from docstore: {len(texts)}")
 
     if not texts:
-        # Write an empty index file so downstream UI can still boot gracefully
-        (INDEX_DIR).mkdir(parents=True, exist_ok=True)
+        INDEX_DIR.mkdir(parents=True, exist_ok=True)
         (INDEX_DIR / "meta.json").write_text(json.dumps([], ensure_ascii=False))
         print("[index] No texts to embed. Wrote empty meta.json.")
         return 0
 
-    # Embed (CPU default; keeps it portable)
+    # Embed (CPU default; portable)
     model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
     model.max_seq_length = 256
-    batch = max(8, min(32, len(texts)))  # sensible batch size for small corpora
+    batch = max(8, min(32, len(texts)))
     emb = model.encode(
         texts,
         convert_to_numpy=True,
@@ -117,22 +368,32 @@ def _build_index_from_docstore() -> int:
     return len(texts)
 
 
-# -------------------- Ingest main --------------------
+# -------------------- Public API: ingest --------------------
+
+__all__ = ["ingest"]
 
 def ingest(cfg_path: str = "config/sources.yaml", env: Dict | None = None):
     """
-    Reads config, fetches from enabled sources, normalizes with a single map,
-    attaches categories/geo consistently, DEDUPEs, and builds the index.
+    Reads config, fetches from enabled sources via adapters, normalizes to a single schema,
+    applies filters (capacity / PA-MD), dedupes, writes docstore, and builds the FAISS index.
+    Returns (docstore_path, n_indexed).
     """
     cfg = load_config(cfg_path)
 
+    # ---- Filters from config ----
+    f_cfg = (cfg or {}).get("filters", {}) or {}
+    capacity_only = bool(f_cfg.get("capacity_only", True))
+    pa_md_only   = bool(f_cfg.get("pa_md_only", False))
+    print(f"[filters] capacity_only = {'TRUE' if capacity_only else 'FALSE'}")
+    print(f"[filters] pa_md_only   = {'TRUE' if pa_md_only else 'FALSE'}")
+
     all_rows: List[Dict[str, Any]] = []
     for entry in cfg.get("sources", []):
         if not entry.get("enabled"):
             continue
 
         name = entry.get("name", "<source>")
-        geo = entry.get("geo") or "US"
+        geo  = entry.get("geo") or "US"
         cats = entry.get("categories") or []
         static = {"geo": geo, "categories": cats}
 
@@ -143,20 +404,37 @@ def ingest(cfg_path: str = "config/sources.yaml", env: Dict | None = None):
             raw_hits = _collect_from_grantsgov_api(entry)
             rows = [normalize("grants_gov", h, static) for h in raw_hits]
 
+        elif typ in ("web_page", "http_html"):
+            rows = _collect_from_http_html(entry, name, static)
+
+        elif typ == "http_pdf":
+            rows = _collect_from_http_pdf(entry, name, static)
+
         elif typ == "local_sample":
             p = Path(entry["path"]).expanduser()
             blob = json.loads(p.read_text(encoding="utf-8"))
             items = blob.get("opportunities") or []
             rows = [normalize("local_sample", op, static) for op in items]
 
-        else:
-            # Future adapters (doj_ojp, state_md, web_page, json_static, …)
-            rows = []
+        # Unknown types => skip silently
+
+        # ---- Apply capacity / geo filters BEFORE collecting ----
+        if rows and (capacity_only or pa_md_only):
+            filtered = []
+            for r in rows:
+                t = _doc_text_from_row(r)
+                if capacity_only and not _is_capacity_building_text(t):
+                    continue
+                if pa_md_only and not _is_pa_md_text(t):
+                    continue
+                filtered.append(r)
+            print(f"[filter] {name}: kept {len(filtered)}/{len(rows)} after filters")
+            rows = filtered
 
         print(f"[collect] {name} → {len(rows)} rows")
         all_rows.extend(rows)
 
-    # ---- DEDUPE (id → url → title) ----
+    # ---- DEDUPE (by id → url → title) ----
     seen, unique = set(), []
     for r in all_rows:
         key = r.get("id") or r.get("url") or r.get("title")
@@ -172,6 +450,8 @@ def ingest(cfg_path: str = "config/sources.yaml", env: Dict | None = None):
     return path, n
 
 
+# -------------------- CLI --------------------
+
 if __name__ == "__main__":
     import argparse
     ap = argparse.ArgumentParser()

config/sources.yaml

@@ -1,4 +1,8 @@
 # Minimal, valid config — v6.3
+filters:
+  capacity_only: true   # keep only capacity-building items
+  pa_md_only: false     # set to true to restrict index to PA/MD
+
 sources:
   # ---------- FEDERAL: Grants.gov (focused for buses/van/mobility & reentry) ----------
 
@@ -83,6 +87,8 @@ sources:
         sortBy: "openDate|desc"
 
   # ---------- STATE & METRO PASS-THROUGHS (FTA 5310 etc.) ----------
+  # NOTE: These require adapters (http_html/web_page/http_pdf) you haven't implemented yet.
+  # They are kept here (enabled) in case your runtime supports them; otherwise set enabled: false.
 
   - name: "Maryland MTA — Grants (incl. 5310)"
     type: web_page
@@ -156,8 +162,149 @@ sources:
       mode: "article"
       keep_links: true
 
-  # ---------- OPTIONAL: Curated JSON (enable after you generate it) ----------
+  # --- Pennsylvania: PCA (state arts) ---
+  - name: "PA Creative Industries – Capacity Building (landing)"
+    type: http_html
+    enabled: true
+    url: "https://www.pa.gov/agencies/coa/grants-and-loans/capacity-building-programs.html"
+    geo: "PA"
+    categories: ["capacity_building"]
+    parse:
+      follow_links: true
+      link_selectors:
+        - "a[href*='capacity']"
+        - "a[href*='strategies-for-success']"
+        - "a[href$='.pdf']"
+      content_selectors:
+        - "main"
+        - "article"
+        - ".content"
+
+  - name: "PA Creative Industries – Creative Sector Flex Fund"
+    type: http_html
+    enabled: true
+    url: "https://www.pa.gov/agencies/coa/grants-and-loans/creative-sector-flex-fund.html"
+    geo: "PA"
+    categories: ["capacity_building"]
+    parse:
+      follow_links: true
+      link_selectors:
+        - "a[href$='.pdf']"
+        - "a[href*='guidelines']"
+        - "a[href*='apply']"
+      content_selectors:
+        - "main"
+        - "article"
+        - ".content"
 
+  # --- Pennsylvania: PCCD (eGrants announcements & PDFs) ---
+  - name: "PCCD – Funding Announcements (eGrants)"
+    type: http_html
+    enabled: true
+    url: "https://www.pccd.pa.gov/Funding/Pages/default.aspx"
+    geo: "PA"
+    categories: ["capacity_building", "public_safety", "youth"]
+    parse:
+      follow_links: true
+      link_selectors:
+        - "a[href*='Funding-Announcement']"
+        - "a[href$='.pdf']"
+        - "a[href*='CJAB']"
+        - "a[href*='VIP']"
+        - "a[href*='CCVI']"
+        - "a[href*='BOOST']"
+      content_selectors:
+        - "main"
+        - "article"
+        - ".ms-rtestate-field"
+
+  - name: "PCCD – PDFs (deep fetch)"
+    type: http_pdf
+    enabled: true
+    url_patterns:
+      - "https://www.pccd.pa.gov/*/*.pdf"
+    geo: "PA"
+    categories: ["capacity_building"]
+
+  # --- Maryland: OneStop (statewide grant listings with 'capacity' search) ---
+  - name: "Maryland OneStop – Capacity search"
+    type: http_html
+    enabled: true
+    url: "https://onestop.md.gov/search?query=capacity"
+    geo: "MD"
+    categories: ["capacity_building"]
+    parse:
+      follow_links: true
+      link_selectors:
+        - "a[href*='/forms/']"
+        - "a[href*='/search/']"
+      content_selectors:
+        - "main"
+        - "article"
+        - "[role='main']"
+
+  # --- Maryland: DHCD (housing/community programs & press) ---
+  - name: "MD DHCD – Programs (grants & loans index)"
+    type: http_html
+    enabled: true
+    url: "https://dhcd.maryland.gov/Pages/Programs.aspx"
+    geo: "MD"
+    categories: ["capacity_building", "housing", "community_development"]
+    parse:
+      follow_links: true
+      link_selectors:
+        - "a[href*='Programs']"
+        - "a[href$='.pdf']"
+        - "a[href*='Trust']"
+      content_selectors:
+        - "#content"
+        - "main"
+        - "article"
+
+  - name: "MD DHCD – Press/Notices (watch for NOFOs)"
+    type: http_html
+    enabled: true
+    url: "https://dhcd.maryland.gov/Pages/PressReleases.aspx"
+    geo: "MD"
+    categories: ["capacity_building"]
+    parse:
+      follow_links: true
+      link_selectors:
+        - "a[href$='.pdf']"
+        - "a[href*='Notice']"
+        - "a[href*='Funding']"
+      content_selectors:
+        - "#content"
+        - "main"
+        - "article"
+
+  # --- Maryland: Chesapeake Bay Trust (recurring capacity-building RFPs) ---
+  - name: "Chesapeake Bay Trust – Capacity Building Initiative (CBI)"
+    type: http_html
+    enabled: true
+    url: "https://cbtrust.org/grants/capacity-building/"
+    geo: "MD"
+    categories: ["capacity_building", "environment", "community_health"]
+    parse:
+      follow_links: true
+      link_selectors:
+        - "a[href$='.pdf']"
+        - "a[href*='Request-for-Proposals']"
+        - "a[href*='RFP']"
+      content_selectors:
+        - "main"
+        - "article"
+        - ".entry-content"
+
+  - name: "CB Trust – PDFs (deep fetch)"
+    type: http_pdf
+    enabled: true
+    url_patterns:
+      - "https://cbtrust.org/*/*.pdf"
+    geo: "MD"
+    categories: ["capacity_building"]
+
+  # ---------- OPTIONAL: Curated JSON (enable after you generate it) ----------
   - name: "State 5310 Listings (curated JSON)"
     type: json_static
     enabled: false     # set to true once you generate the file below

project-plan-rag.rtf

@@ -0,0 +1,569 @@
+{\rtf1\ansi\ansicpg1252\cocoartf2822
+\cocoatextscaling0\cocoaplatform0{\fonttbl\f0\froman\fcharset0 Times-Bold;\f1\froman\fcharset0 Times-Roman;\f2\fmodern\fcharset0 Courier;
+\f3\froman\fcharset0 TimesNewRomanPSMT;\f4\fnil\fcharset0 AppleColorEmoji;\f5\froman\fcharset0 Times-Italic;
+\f6\fnil\fcharset0 Menlo-Regular;\f7\fnil\fcharset0 HelveticaNeue;}
+{\colortbl;\red255\green255\blue255;\red0\green0\blue0;\red0\green0\blue233;\red109\green109\blue109;
+\red109\green109\blue109;\red0\green0\blue0;}
+{\*\expandedcolortbl;;\cssrgb\c0\c0\c0;\cssrgb\c0\c0\c93333;\cssrgb\c50196\c50196\c50196;
+\cssrgb\c50196\c50196\c50196;\cssrgb\c0\c0\c0\c84706;}
+{\*\listtable{\list\listtemplateid1\listhybrid{\listlevel\levelnfc23\levelnfcn23\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{disc\}}{\leveltext\leveltemplateid1\'01\uc0\u8226 ;}{\levelnumbers;}\fi-360\li720\lin720 }{\listname ;}\listid1}
+{\list\listtemplateid2\listhybrid{\listlevel\levelnfc0\levelnfcn0\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{decimal\}}{\leveltext\leveltemplateid101\'01\'00;}{\levelnumbers\'01;}\fi-360\li720\lin720 }{\listlevel\levelnfc23\levelnfcn23\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{circle\}}{\leveltext\leveltemplateid102\'01\uc0\u9702 ;}{\levelnumbers;}\fi-360\li1440\lin1440 }{\listname ;}\listid2}
+{\list\listtemplateid3\listhybrid{\listlevel\levelnfc23\levelnfcn23\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{disc\}}{\leveltext\leveltemplateid201\'01\uc0\u8226 ;}{\levelnumbers;}\fi-360\li720\lin720 }{\listname ;}\listid3}
+{\list\listtemplateid4\listhybrid{\listlevel\levelnfc0\levelnfcn0\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{decimal\}}{\leveltext\leveltemplateid301\'01\'00;}{\levelnumbers\'01;}\fi-360\li720\lin720 }{\listlevel\levelnfc23\levelnfcn23\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{circle\}}{\leveltext\leveltemplateid302\'01\uc0\u9702 ;}{\levelnumbers;}\fi-360\li1440\lin1440 }{\listlevel\levelnfc23\levelnfcn23\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{square\}}{\leveltext\leveltemplateid303\'01\uc0\u9642 ;}{\levelnumbers;}\fi-360\li2160\lin2160 }{\listname ;}\listid4}
+{\list\listtemplateid5\listhybrid{\listlevel\levelnfc23\levelnfcn23\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{disc\}}{\leveltext\leveltemplateid401\'01\uc0\u8226 ;}{\levelnumbers;}\fi-360\li720\lin720 }{\listname ;}\listid5}
+{\list\listtemplateid6\listhybrid{\listlevel\levelnfc0\levelnfcn0\leveljc0\leveljcn0\levelfollow0\levelstartat1\levelspace360\levelindent0{\*\levelmarker \{decimal\}}{\leveltext\leveltemplateid501\'01\'00;}{\levelnumbers\'01;}\fi-360\li720\lin720 }{\listname ;}\listid6}}
+{\*\listoverridetable{\listoverride\listid1\listoverridecount0\ls1}{\listoverride\listid2\listoverridecount0\ls2}{\listoverride\listid3\listoverridecount0\ls3}{\listoverride\listid4\listoverridecount0\ls4}{\listoverride\listid5\listoverridecount0\ls5}{\listoverride\listid6\listoverridecount0\ls6}}
+\margl1440\margr1440\vieww31340\viewh19300\viewkind0
+\deftab720
+\pard\pardeftab720\sa298\partightenfactor0
+
+\f0\b\fs36 \cf0 \expnd0\expndtw0\kerning0
+Best practices & features to include\
+\pard\pardeftab720\sa240\partightenfactor0
+
+\f1\b0\fs24 \cf0 From the literature + what you\'92ve built already, here are features that improve quality & usability. {\field{\*\fldinst{HYPERLINK "https://www.funraise.org/blog/grant-management-software-for-nonprofits?utm_source=chatgpt.com"}}{\fldrslt \cf3 \ul \ulc3 NetSuite+3Funraise+3Fluxx+3}}\
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrt\brdrnil \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\qc\partightenfactor0
+
+\f0\b \cf0 Feature\cell 
+\pard\intbl\itap1\pardeftab720\qc\partightenfactor0
+\cf0 Why it matters\cell 
+\pard\intbl\itap1\pardeftab720\qc\partightenfactor0
+\cf0 How to implement / what to watch out for\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Opportunity filters
+\f1\b0  (keywords, geography, type, capacity-building etc.)\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Helps users narrow to what matters, reduces overload.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 You have keyword filters already. Also include date, state, amount range, \'93open vs closed\'94 status.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 Deadline alerts / reminders
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Prevents missing grants because deadlines slip.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Track 
+\f2\fs26 deadline
+\f1\fs24  (if parsed), then show upcoming ones. Allow export to calendar or reminders.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 Document / PDF fetching + parsing
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Many state sources have PDFs, RFPs etc. Users want details, not just summary.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Use PDF adapter + follow-links. Flag PDFs clearly in UI.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 Staleness / expiration detection
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Projects with expired deadlines clutter feeds.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Tag items with \'93deadline passed\'94 or \'93pending\'94 etc. Maybe auto-hide old ones after some time.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 User feedback / manual review / save items
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Users can flag false positives, save promising ones.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Add \'93bookmark / save\'94 or \'93dismiss\'94 features. Could feed into machine learning or heuristics over time.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 Dashboard / Analytics
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Helps see grant volume by type, deadlines, states, etc.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Simple charts like # grants by month, # capacity-building grants vs total, etc.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 Good defaults / simple UI
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Avoid cognitive overload \'97 show only essentials, allow advanced filtering if needed.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 E.g. show title, deadline, source, eligibility; hide long descriptions by default.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 Mobile-friendly / responsive
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Many will check on phones or tablets.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 If web UI, ensure collapsible fields, simple menus.\cell \row
+
+\itap1\trowd \taflags0 \trgaph108\trleft-108 \trbrdrl\brdrnil \trbrdrt\brdrnil \trbrdrr\brdrnil 
+\clvertalc \clshdrawnil \clwWidth4256\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx2880
+\clvertalc \clshdrawnil \clwWidth5037\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx5760
+\clvertalc \clshdrawnil \clwWidth6106\clftsWidth3 \clmart10 \clmarl10 \clmarb10 \clmarr10 \clbrdrt\brdrnil \clbrdrl\brdrnil \clbrdrb\brdrnil \clbrdrr\brdrnil \clpadt20 \clpadl20 \clpadb20 \clpadr20 \gaph\cellx8640
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+
+\f0\b \cf0 Integrations
+\f1\b0 \cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 E.g. calendar, Slack/email alerts.\cell 
+\pard\intbl\itap1\pardeftab720\partightenfactor0
+\cf0 Helps push notifications rather than only manual checking.\cell \lastrow\row
+\pard\pardeftab720\partightenfactor0
+\cf4 \
+\pard\pardeftab720\sa298\partightenfactor0
+
+\f0\b\fs36 \cf0 What to avoid / limit to keep usability\
+\pard\pardeftab720\sa240\partightenfactor0
+
+\f1\b0\fs24 \cf0 These things often add data/noise or complexity, so either avoid or gate them behind \'93advanced\'94 toggles.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls1\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Over-broad scraping
+\f1\b0 : pulling every \'93policy\'94, \'93program info\'94, \'93newsletter\'94, etc. just because capacity keywords appear once.\
+\ls1\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Too much automation without oversight
+\f1\b0 : e.g. hidden deadlines if PDF text parsing fails.\
+\ls1\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Overwhelming metadata fields
+\f1\b0 : stuff like \'93agency budget history\'94, \'93application score weights\'94 etc unless users request them.\
+\ls1\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Frequent UI changes or too many fields
+\f1\b0  \'97 keep interface consistent.\
+\ls1\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Large batch of false positives
+\f1\b0  \'97 this ruins trust quickly.\
+\pard\pardeftab720\partightenfactor0
+\cf4 \
+\pard\pardeftab720\sa298\partightenfactor0
+
+\f0\b\fs36 \cf0 What to build next in your tool\
+\pard\pardeftab720\sa240\partightenfactor0
+
+\f1\b0\fs24 \cf0 Here are prioritized improvements/next features for your tool to make it more powerful while preserving clarity:\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	1	}\expnd0\expndtw0\kerning0
+Deadline extraction / detection
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+From Grants.gov API: likely available.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+From scraped state pages / PDFs: attempt to parse \'93deadline\'94, \'93closing date\'94. If missing, mark as \'93TBD\'94.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+UI: highlight upcoming deadlines (\'93Due in next 30 days\'94).\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	2	}\expnd0\expndtw0\kerning0
+\'93Open / Active\'94 flag
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+If RFPs are active or you can detect \'93application now open\'94 in text.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+If not, mark as \'93reference / program\'94 so users know it\'92s structural info, not a live call.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	3	}\expnd0\expndtw0\kerning0
+Bookmark / dismiss / feedback
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Users should be able to mark \'93this is useful\'94 or \'93not relevant\'94 to train future filtering.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Possibly store local tags (e.g. \'93my state\'94, \'93my priority\'94).\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	4	}\expnd0\expndtw0\kerning0
+Improved link / PDF following
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+As suggested earlier.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Make sure link selectors are fine-tuned for each state source.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	5	}\expnd0\expndtw0\kerning0
+Expired / historical items archive
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Hide automatically unless explicitly requested. Keeps main view clean.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	6	}\expnd0\expndtw0\kerning0
+Search + sort + filter UI in front end
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Filters: State, source type (federal/state), capacity vs other, keyword.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Sort by deadline, date posted, amount (if available).\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	7	}\expnd0\expndtw0\kerning0
+Notifications / reminders
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Let users get alerts (email / Slack / calendar) for items that match their saved filters and have upcoming deadlines.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	8	}\expnd0\expndtw0\kerning0
+Confidence score or matching hint
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+For scraped items, display \'93match strength\'94 (how many capacity keywords matched, whether in title vs body). Helps users see which items are likely relevant.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	9	}\expnd0\expndtw0\kerning0
+Performance / scheduling
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Check feeds regularly, avoid stale caches.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Maybe incremental ingest (only new items) rather than full crawl all the time.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls2\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	10	}\expnd0\expndtw0\kerning0
+Testing & feedback
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls2\ilvl1\cf0 \kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Ask actual users (your team) to test beta versions and tell you what\'92s too much / too little.\
+\ls2\ilvl1\kerning1\expnd0\expndtw0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+Adjust based on real use.\
+\pard\pardeftab720\partightenfactor0
+\cf4 \
+\pard\pardeftab720\sa298\partightenfactor0
+
+\f0\b\fs36 \cf0 Your roadmap can look like this\
+\pard\pardeftab720\sa240\partightenfactor0
+
+\f1\b0\fs24 \cf0 Here\'92s a potential plan for the next few sprints:\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls3\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Sprint 1
+\f1\b0  (now): Add deadline extraction + \'93active\'94 flag + improved link following for state sources.\
+\ls3\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Sprint 2
+\f1\b0 : Bookmark/dismiss features + advanced filtering UI + sort by deadline.\
+\ls3\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Sprint 3
+\f1\b0 : Notifications/reminders + mobile UI polish + confidence scoring.\
+\ls3\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+Sprint 4
+\f1\b0 : Incorporate JS-rendered sources (Playwright) + archive/historical mode.\
+\pard\tx720\pardeftab720\sa240\partightenfactor0
+\cf0 \
+\pard\pardeftab720\sa240\partightenfactor0
+\cf0 \outl0\strokewidth0 \strokec2 Yes\'97exactly. We can treat the roadmap I outlined earlier as 
+\f0\b four sprints
+\f1\b0 .\uc0\u8232 Here\'92s a clear 
+\f0\b Sprint 1 plan
+\f1\b0  (2\'963 weeks is typical) so we can start delivering value quickly while keeping the other sprints in view.\
+\pard\pardeftab720\partightenfactor0
+\cf5 \strokec5 \
+\pard\pardeftab720\sa298\partightenfactor0
+
+\f4\fs36 \cf0 \strokec2 \uc0\u55356 \u57281 
+\f0\b  Sprint 1 \'96 \'93Actionable Deadlines & Better State Data\'94\
+\pard\pardeftab720\sa240\partightenfactor0
+
+\fs24 \cf0 Goal:
+\f1\b0  Surface 
+\f5\i current
+\f1\i0  and 
+\f5\i timely
+\f1\i0  grant opportunities\'97especially from PA & MD\'97without flooding users.\
+\pard\pardeftab720\sa280\partightenfactor0
+
+\f0\b\fs28 \cf0 Deliverables\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls4\ilvl0
+\fs24 \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	1	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Deadline Extraction & \'93Active\'94 Flag
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls4\ilvl1\cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Parse 
+\f2\fs26 deadline
+\f1\fs24 /
+\f2\fs26 closing date
+\f1\fs24  text from Grants.gov (API gives it directly).\
+\ls4\ilvl1\kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Add a lightweight NLP/regex extractor for state HTML/PDFs (\'93Deadline\'94, \'93Applications due\'94, etc.).\
+\ls4\ilvl1\kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 In the index, add fields: 
+\f2\fs26 deadline_date
+\f1\fs24 , 
+\f2\fs26 is_active
+\f1\fs24  (true if deadline >= today or marked open/TBD).\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls4\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	2	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 UI Updates in the Dashboard
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls4\ilvl1\cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Show deadlines and days-to-close (already in the mock).\
+\ls4\ilvl1\kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Add a small 
+\f0\b \'93Active Only\'94
+\f1\b0  toggle to hide expired or guideline-only records.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls4\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	3	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Link-Following Enhancements
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls4\ilvl1\cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Update YAML for 
+\f0\b MD DHCD Press
+\f1\b0 , 
+\f0\b MTA
+\f1\b0 , 
+\f0\b PennDOT
+\f1\b0 , etc. to:\
+\pard\tx1660\tx2160\pardeftab720\li2160\fi-2160\sa240\partightenfactor0
+\ls4\ilvl2
+\f2\fs26 \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f6 \uc0\u9642 
+\f2 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 parse.follow_links: true
+\f1\fs24 \
+\ls4\ilvl2
+\f2\fs26 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f6 \uc0\u9642 
+\f2 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 crawl.max_depth: 1
+\f1\fs24 \
+\ls4\ilvl2
+\f2\fs26 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f6 \uc0\u9642 
+\f2 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 link_selectors
+\f1\fs24  for PDFs / NOFO pages\
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls4\ilvl1\cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Ensure adapter fetches and normalizes these linked pages.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls4\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	4	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Improved Capacity Filter
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls4\ilvl1\cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Add a few transportation-related capacity phrases (\'93service capacity\'94, \'93provider capacity\'94) to capture more relevant 5310-type opportunities.\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls4\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	5	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Basic Quality Checks
+\f1\b0 \
+\pard\tx940\tx1440\pardeftab720\li1440\fi-1440\sa240\partightenfactor0
+\ls4\ilvl1\cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Logging: show counts of items skipped due to expired deadlines or missing text.\
+\ls4\ilvl1\kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	
+\f3 \uc0\u9702 
+\f1 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Quick manual review with your team to confirm that the new records are genuinely grant opportunities.\
+\pard\pardeftab720\partightenfactor0
+\cf5 \strokec5 \
+\pard\pardeftab720\sa298\partightenfactor0
+
+\f4\fs36 \cf0 \strokec2 \uc0\u55357 \u56621 
+\f0\b  Sprint 2\'964 Preview\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls5\ilvl0
+\fs24 \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Sprint 2:
+\f1\b0  Bookmarks & Dismiss, Advanced Filtering UI, Sorting by deadline/confidence.\
+\ls5\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Sprint 3:
+\f1\b0  Notifications (email/Slack/ICS calendar) + mobile UI polish + confidence scoring display.\
+\ls5\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	\uc0\u8226 	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Sprint 4:
+\f1\b0  Full 
+\f0\b JS-rendered
+\f1\b0  source support (Playwright for MD OneStop etc.) and historical archive mode.\
+\pard\pardeftab720\partightenfactor0
+\cf5 \strokec5 \
+\pard\pardeftab720\sa298\partightenfactor0
+
+\f0\b\fs36 \cf0 \strokec2 Next Steps to Kick Off Sprint 1\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls6\ilvl0
+\fs24 \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	1	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Confirm dev environment
+\f1\b0  \'96 you already have 
+\f2\fs26 grants-rag
+\f1\fs24  running locally.\
+\ls6\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	2	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Create a new branch
+\f1\b0 :\uc0\u8232 
+\f2\fs26 \uc0\u8232 \u8232 \u8232 git checkout -b sprint1-deadline-active-flag\
+\pard\tx220\tx720\pardeftab720\li720\fi-720\partightenfactor0
+\ls6\ilvl0\cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	3	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 \uc0\u8232 \u8232 
+\f1\fs24 \
+\pard\tx220\tx720\pardeftab720\li720\fi-720\sa240\partightenfactor0
+\ls6\ilvl0
+\f0\b \cf0 \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	4	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 Add extraction utilities
+\f1\b0 \uc0\u8232 
+\f5\i Regex for state HTML/PDFs
+\f1\i0  and update 
+\f2\fs26 _normalize_web_record
+\f1\fs24  to accept 
+\f2\fs26 deadline_date
+\f1\fs24  and 
+\f2\fs26 is_active
+\f1\fs24 .\
+\ls6\ilvl0
+\f0\b \kerning1\expnd0\expndtw0 \outl0\strokewidth0 {\listtext	5	}\expnd0\expndtw0\kerning0
+\outl0\strokewidth0 \strokec2 UI
+\f1\b0 : enable the 
+\f0\b Active Only
+\f1\b0  toggle (mock is already designed for it).\
+\pard\pardeftab720\partightenfactor0
+\cf5 \strokec5 \
+\pard\pardeftab720\sa240\partightenfactor0
+
+\f0\b \cf0 \strokec2 Timeline suggestion:
+\f1\b0  2 weeks dev + 1 week QA / user feedback.\uc0\u8232 Once Sprint 1 is merged, we\'92ll have a tool that surfaces only 
+\f5\i active
+\f1\i0  capacity-building grants with clear deadlines\'97exactly the balance of \'93rich data without overload\'94 we\'92re after.\
+\pard\pardeftab720\partightenfactor0
+\cf0 \
+\
+\
+\
+\
+\
+\
+\pard\pardeftab720\sa240\partightenfactor0
+\cf0 \
+\pard\pardeftab720\partightenfactor0
+\cf0 \
+\pard\pardeftab720\qc\partightenfactor0
+
+\f7\fs22 \cf6 \strokec6 \
+\pard\pardeftab720\partightenfactor0
+
+\f1\fs24 \cf0 \strokec2 ChatGPT can make mistakes. Check important info.\
+\pard\tx720\pardeftab720\sa240\partightenfactor0
+\cf0 \outl0\strokewidth0 \
+}