#!/usr/bin/env python3
"""kb_graph.py — đồ thị tri thức company_kb (kiểu Obsidian) cho dashboard. CHỈ-ĐỌC.

nodes = mỗi file .md trong company_kb; edges từ 3 nguồn:
  1) [[wikilink]] giữa note,
  2) cấu trúc: business/<fw>/<block> → node framework <fw> (gom cụm canvas),
  3) #tag chung (nối các note cùng tag, bỏ tag quá phổ biến).

CLI:  python3 lib/kb_graph.py graph        -> {ok, nodes, edges, stats}
      python3 lib/kb_graph.py note <id>    -> {ok, id, title, body}
"""
import os
import re
import sys
import json


def _bundle_root():
    # Cập nhật đường dẫn gốc về workspace của CEO để tìm thấy company_kb
    # File hiện tại: /opt/ai-os/core/lib/kb_graph.py
    # Gốc cần trỏ về: /opt/ai-os/products/ceo
    # Thay vì lùi 2 cấp về /opt/ai-os/core, ta lùi 3 cấp rồi vào products/ceo
    core_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
    ai_os_dir = os.path.dirname(core_dir)
    return os.path.join(ai_os_dir, "products", "ceo")


# Global variables and configuration
ROOT = _bundle_root()
KB = os.path.join(ROOT, "company_kb")
SKIP = {"readme", "_index", "_profile"}          # file meta/sinh tự động — bỏ khỏi graph

WIKILINK = re.compile(r"\[\[([^\]\|#]+)(?:[#\|][^\]]*)?\]\]")
MDLINK = re.compile(r"\[[^\]]*\]\(([^)\s]+\.md)\)")        # OKF: [text](path.md)
TAG = re.compile(r"(?:^|\s)#([A-Za-z][A-Za-z0-9_\-/]+)")


def _rel_id(path):
    rel = os.path.relpath(path, KB).replace(os.sep, "/")
    return rel[:-3] if rel.endswith(".md") else rel


def _title(path, nid):
    try:
        for ln in open(path, encoding="utf-8"):
            s = ln.strip()
            if s.startswith("#"):
                return s.lstrip("# ").strip() or nid.split("/")[-1]
    except Exception:
        pass
    return nid.split("/")[-1]


def _body(path):
    try:
        t = open(path, encoding="utf-8").read()
    except Exception:
        return ""
    if t.startswith("---"):
        p = t.split("---", 2)
        if len(p) == 3:
            t = p[2]
    return t.lstrip("\n")


def _walk():
    if not os.path.isdir(KB):
        return
    for dp, _, fns in os.walk(KB):
        for fn in fns:
            if fn.endswith(".md"):
                yield os.path.join(dp, fn)


def build():
    nodes, by_base, raw = {}, {}, {}
    for f in _walk():
        nid = _rel_id(f)
        base = nid.split("/")[-1].lower()
        if base in SKIP:
            continue
            
        # Exclude raw notebooklm_ingest files to keep graph clean
        if nid.startswith("notebooklm_ingest/"):
            continue
            
        # Group by sub-categories under knowledge/
        parts = nid.split("/")
        if len(parts) >= 2 and parts[0] == "knowledge":
            grp = parts[1]
        else:
            grp = parts[0] if "/" in nid else "note"
            
        nodes[nid] = {"id": nid, "label": _title(f, nid), "group": grp, "kind": "note", "deg": 0}
        by_base.setdefault(base, nid)
        raw[nid] = _body(f)

    edges, seen = [], set()

    def add_edge(s, t, kind):
        if s == t or s not in nodes or t not in nodes:
            return
        k = tuple(sorted((s, t))) + (kind,)
        if k in seen:
            return
        seen.add(k)
        edges.append({"s": s, "t": t, "kind": kind})
        nodes[s]["deg"] += 1
        nodes[t]["deg"] += 1

    # 1) liên kết — Obsidian [[wikilink]] + OKF [text](path.md). Path = danh tính.
    def _resolve(tgt):
        tgt = tgt.strip().split("#")[0].split("|")[0].strip().strip("/")
        tgt = re.sub(r"\.md$", "", tgt)
        if not tgt:
            return None
        if tgt in nodes:                       # khớp path đầy đủ (vd business/bmc/customers)
            return tgt
        return by_base.get(tgt.split("/")[-1].lower())   # hoặc theo basename
        
    for nid, body in raw.items():
        for m in list(WIKILINK.findall(body)) + list(MDLINK.findall(body)):
            t = _resolve(m)
            if t:
                add_edge(nid, t, "link")

    # 2) cấu trúc business/<fw>/<block> → node framework
    fw_new = {}
    for nid in list(nodes):
        p = nid.split("/")
        if len(p) >= 3 and p[0] == "business":
            fwid = "business/%s" % p[1]
            if fwid not in nodes:
                fw_new[fwid] = {"id": fwid, "label": p[1], "group": "framework", "kind": "framework", "deg": 0}
    nodes.update(fw_new)
    for nid in list(nodes):
        p = nid.split("/")
        if len(p) >= 3 and p[0] == "business":
            add_edge(nid, "business/%s" % p[1], "structure")

    # 3) tag chung
    tag_map = {}
    for nid, body in raw.items():
        for tg in set(TAG.findall(body)):
            tag_map.setdefault(tg.lower(), []).append(nid)
    for ids in tag_map.values():
        if 2 <= len(ids) <= 12:
            for i in range(len(ids)):
                for j in range(i + 1, len(ids)):
                    add_edge(ids[i], ids[j], "tag")

    ns = list(nodes.values())
    return {"ok": True, "nodes": ns, "edges": edges,
            "stats": {"notes": sum(1 for n in ns if n["kind"] == "note"),
                      "frameworks": sum(1 for n in ns if n["kind"] == "framework"),
                      "edges": len(edges)}}


def note(nid):
    rel = re.sub(r"\.\.+", "", (nid or "").strip().strip("/"))
    target = os.path.realpath(os.path.join(KB, rel + ".md"))
    base = os.path.realpath(KB)
    if not (target == base or target.startswith(base + os.sep)) or not os.path.isfile(target):
        return {"ok": False, "error": "not_found"}
    rid = _rel_id(target)
    return {"ok": True, "id": rid, "title": _title(target, rid), "body": _body(target).strip()}


def main():
    a = sys.argv[1:]
    cmd = a[0] if a else "graph"
    if cmd == "graph":
        out = build()
    elif cmd == "note":
        out = note(a[1] if len(a) > 1 else "")
    else:
        out = {"ok": False, "error": "unknown cmd"}
    print(json.dumps(out, ensure_ascii=False))
    return 0


if __name__ == "__main__":
    sys.exit(main())