diff --git a/proxmox/lxc171-claude-agent/discord-bridge/dashboard/api.py b/proxmox/lxc171-claude-agent/discord-bridge/dashboard/api.py
index f13ee01..e494bdc 100644
--- a/proxmox/lxc171-claude-agent/discord-bridge/dashboard/api.py
+++ b/proxmox/lxc171-claude-agent/discord-bridge/dashboard/api.py
@@ -61,6 +61,9 @@ MARIA_DOC_EXTENSIONS = (".txt", ".md", ".xml")
_MARIA_FORMAT_RANK = {".xml": 0, ".md": 1, ".txt": 2}
MARIA_STATE_DIR = Path(os.environ.get("MARIA_BRIDGE_DIR") or (Path.home() / ".maria-bridge"))
MARIA_DOCS_DIR = MARIA_STATE_DIR / "documents"
+# Documentele care nu vin din Drive (rclone sync sterge din MARIA_DOCS_DIR tot ce
+# nu exista in Drive). Oglinda lui rag/config.py:DOCS_LOCAL_DIR.
+MARIA_DOCS_LOCAL_DIR = MARIA_STATE_DIR / "documents-local"
MARIA_INDEX_FILE = MARIA_STATE_DIR / "rag_index.json"
MARIA_ENV_FILE = MARIA_STATE_DIR / "env"
MARIA_LOG_DIR = MARIA_STATE_DIR / "logs"
@@ -144,10 +147,14 @@ def maria_documents() -> list[dict]:
Nu importam modulul acela: si el, si `discord-bridge/config.py` s-ar numi `config`.
"""
MARIA_DOCS_DIR.mkdir(parents=True, exist_ok=True)
- files = [
- f for f in sorted(MARIA_DOCS_DIR.glob("*"))
- if f.is_file() and f.suffix in MARIA_DOC_EXTENSIONS
- ]
+ vazute: dict[str, Path] = {}
+ for d in (MARIA_DOCS_LOCAL_DIR, MARIA_DOCS_DIR): # Drive ultimul = castiga la acelasi nume
+ if not d.is_dir():
+ continue
+ for f in sorted(d.glob("*")):
+ if f.is_file() and f.suffix in MARIA_DOC_EXTENSIONS:
+ vazute[f.name] = f
+ files = [vazute[n] for n in sorted(vazute)]
winners: dict[str, str] = {}
for f in files:
best = winners.get(f.stem)
@@ -162,6 +169,7 @@ def maria_documents() -> list[dict]:
"name": f.name,
"size": st.st_size,
"mtime": st.st_mtime,
+ "local": f.parent == MARIA_DOCS_LOCAL_DIR,
"shadowed_by": None if winner == f.name else winner,
})
return out
@@ -206,18 +214,29 @@ def _maria_validate_name(name: str) -> str:
def maria_write_document(name: str, content: str) -> None:
+ """Scrie in documents-local/, nu in oglinda Drive.
+
+ Pana acum ajungea in `documents/` si disparea la urmatorul `rclone sync` —
+ documentul adaugat din dashboard traia pana la 10 minute.
+ """
name = _maria_validate_name(name)
- MARIA_DOCS_DIR.mkdir(parents=True, exist_ok=True)
- (MARIA_DOCS_DIR / name).write_text(content, encoding="utf-8")
+ MARIA_DOCS_LOCAL_DIR.mkdir(parents=True, exist_ok=True)
+ (MARIA_DOCS_LOCAL_DIR / name).write_text(content, encoding="utf-8")
def maria_delete_document(name: str) -> bool:
+ """Sterge documentul din oricare din cele doua directoare.
+
+ Cel din Drive revine la urmatoarea sincronizare — sterge-l din Drive daca vrei
+ sa dispara definitiv.
+ """
name = _maria_validate_name(name)
- path = MARIA_DOCS_DIR / name
- if not path.exists():
- return False
- path.unlink()
- return True
+ for d in (MARIA_DOCS_DIR, MARIA_DOCS_LOCAL_DIR):
+ path = d / name
+ if path.exists():
+ path.unlink()
+ return True
+ return False
def _maria_run(script: str, timeout: float = 600.0, extra_args: list[str] | None = None) -> dict:
diff --git a/proxmox/lxc171-claude-agent/discord-bridge/dashboard/index.html b/proxmox/lxc171-claude-agent/discord-bridge/dashboard/index.html
index 16f8904..870d6df 100644
--- a/proxmox/lxc171-claude-agent/discord-bridge/dashboard/index.html
+++ b/proxmox/lxc171-claude-agent/discord-bridge/dashboard/index.html
@@ -466,10 +466,11 @@ async function mRefresh() {
document.getElementById('mDocCount').textContent =
docs.length + ' în depozit' + (indexate !== docs.length ? ' · ' + indexate + ' indexate' : '');
document.getElementById('mDocs').innerHTML = docs.length ? docs.map(function (d) {
+ var local = d.local ? ' · local' : '';
var umbrit = d.shadowed_by
? ' · umbrit de ' + esc(d.shadowed_by) + ''
: '';
- return '
' + esc(d.name) + umbrit + '
' +
+ return '
' + esc(d.name) + local + umbrit + '
' +
'
' + d.size + ' B
' +
'
';
}).join('') : '
niciun document
';
diff --git a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/README.md b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/README.md
index 6a8944b..3f0df5a 100644
--- a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/README.md
+++ b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/README.md
@@ -197,10 +197,21 @@ puntii Discord — vezi `../discord-bridge/README.md` pentru URL si autentificar
## Depozitul de documente
-Fisiere `.txt`/`.md`/`.xml` in `~/.maria-bridge/documents/` (lista exacta:
-`store.DOC_EXTENSIONS`). Se pot administra:
-1. **manual din dashboard-ul comun** (adaugare/stergere text, reindexare automata la salvare);
-2. **prin sincronizare Google Drive** — vezi mai jos.
+Fisiere `.txt`/`.md`/`.xml` (lista exacta: `store.DOC_EXTENSIONS`), in **doua
+directoare care formeaza un singur spatiu de nume**:
+
+| Director | Ce e acolo |
+|---|---|
+| `~/.maria-bridge/documents/` | **oglinda Google Drive.** `rclone sync` sterge de aici tot ce nu exista in Drive — nu pune nimic de mana. |
+| `~/.maria-bridge/documents-local/` | documente care **nu** vin din Drive: dictionare proprii, ce se adauga din dashboard. Sincronizarea nu se atinge de ele. |
+
+Cand acelasi nume apare in ambele, castiga cel din Drive (e sursa comuna a echipei;
+copia locala ar putea fi o versiune veche uitata acolo). In dashboard, documentele
+locale sunt marcate „local".
+
+Directorul local exista fiindca altfel **tot ce se adauga din dashboard traia pana
+la urmatorul tur de sincronizare** — maximum 10 minute. Acum adaugarea din dashboard
+scrie direct acolo.
## Sincronizare cu Google Drive
@@ -245,9 +256,9 @@ esuata — reia pasii 1-2.
Restul din dosar — chatflow-uri Flowise `.json`, scripturi `.ps1`, `.docx` — sunt
ignorate deliberat: nu sunt cunostinte de suport.
-**`sync` sterge local ce nu mai exista in Drive**, deci depozitul e o oglinda a
-dosarului din Drive, nu o colectie care creste. Documentele adaugate manual din
-dashboard dispar la prima sincronizare daca nu exista si in Drive.
+**`sync` sterge local ce nu mai exista in Drive**, deci `documents/` e o oglinda a
+dosarului din Drive, nu o colectie care creste. De aceea exista `documents-local/`:
+ce e adaugat de mana sau din dashboard sta acolo si nu e atins de sincronizare.
### Dictionarul de erori Oracle
@@ -260,10 +271,13 @@ din captura.
Doua lucruri de stiut:
-- **E samanta, nu sursa.** Fisierul din git nu ajunge singur in index. Depozitul e
- o oglinda a Drive-ului, deci copia pusa direct in `~/.maria-bridge/documents/`
- **dispare la prima sincronizare**. Ca sa ramana, pune-l in `document_store` din
- Drive (pe Windows: `D:\GoogleDrive\romfast\document_store`).
+- **Fisierul din git nu ajunge singur in index.** Copia care ruleaza acum sta in
+ `~/.maria-bridge/documents-local/`, in afara oglinzii Drive. Asa supravietuieste
+ sincronizarilor, dar **exista doar pe acest container**. Locul lui pe termen lung
+ e `document_store` din Drive (pe Windows: `D:\GoogleDrive\romfast\document_store`),
+ unde il vede si restul echipei si de unde se poate edita fara acces la server.
+ Cand ajunge in Drive, sterge copia locala — altfel raman doua versiuni si castiga
+ tacut cea din Drive.
- **E scris pentru clienti, nu pentru administratori.** Nicio rezolvare nu contine
nume de servere, IP-uri, porturi sau pasi de administrare — la erorile care chiar
se rezolva doar tehnic (ORA-01652 spatiu plin, ORA-28000 cont blocat) textul spune
diff --git a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/ops/calibrate-rank.py b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/ops/calibrate-rank.py
index 48273a3..1acd0e3 100755
--- a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/ops/calibrate-rank.py
+++ b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/ops/calibrate-rank.py
@@ -44,6 +44,11 @@ CAZURI: list[tuple[str, bool]] = [
("cum inregistrez o factura de achizitie in ROA", True),
("bilant contabil", True),
("import de date", True),
+ # dupa adaugarea dictionarului de erori Oracle (oracle-erori-uzuale.xml)
+ ("ORA-01722 invalid number la salvarea facturii", True),
+ ("imi da ORA-12154 cand deschid programul", True),
+ ("ORA-00001 unique constraint violated", True),
+ ("eroare ORA-28000 the account is locked", True),
("care e capitala Frantei", False),
("cat costa un bilet de avion la Paris", False),
("imi da eroare la imprimanta HP LaserJet, nu trage hartia", False),
diff --git a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/config.py b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/config.py
index 4a6cdae..ee1ee24 100644
--- a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/config.py
+++ b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/config.py
@@ -17,6 +17,10 @@ _DEFAULT_DIR = pathlib.Path.home() / ".maria-bridge"
STATE_DIR: pathlib.Path = pathlib.Path(os.environ.get("MARIA_BRIDGE_DIR") or _DEFAULT_DIR)
DOCS_DIR: pathlib.Path = STATE_DIR / "documents"
+# Documente care NU vin din Drive. `rclone sync` sterge din DOCS_DIR tot ce nu
+# exista in Drive, deci orice fisier pus acolo de mana dispare la urmatorul tur.
+# Ce sta aici e in afara oglinzii si supravietuieste.
+DOCS_LOCAL_DIR: pathlib.Path = STATE_DIR / "documents-local"
INDEX_FILE: pathlib.Path = STATE_DIR / "rag_index.json"
LOCK_FILE: pathlib.Path = STATE_DIR / ".rag.lock"
LOG_DIR: pathlib.Path = STATE_DIR / "logs"
@@ -87,11 +91,12 @@ def parse_env(text: str) -> dict[str, str]:
def reload(base_dir: str | os.PathLike | None = None) -> dict[str, str]:
"""Recalculeaza caile si reciteste env-ul. Returneaza dictionarul incarcat."""
- global STATE_DIR, DOCS_DIR, INDEX_FILE, LOCK_FILE, LOG_DIR, ENV_FILE, AUTH_DIR, _env
+ global STATE_DIR, DOCS_DIR, DOCS_LOCAL_DIR, INDEX_FILE, LOCK_FILE, LOG_DIR, ENV_FILE, AUTH_DIR, _env
if base_dir is None:
base_dir = os.environ.get("MARIA_BRIDGE_DIR") or _DEFAULT_DIR
STATE_DIR = pathlib.Path(base_dir)
DOCS_DIR = STATE_DIR / "documents"
+ DOCS_LOCAL_DIR = STATE_DIR / "documents-local"
INDEX_FILE = STATE_DIR / "rag_index.json"
LOCK_FILE = STATE_DIR / ".rag.lock"
LOG_DIR = STATE_DIR / "logs"
diff --git a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/store.py b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/store.py
index 5cfb2cb..bab27f0 100644
--- a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/store.py
+++ b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/store.py
@@ -4,6 +4,14 @@ Fisiere text (.txt/.md/.xml) sub STATE_DIR/documents/, un singur nivel (fara
subdirectoare), ca numele afisat in dashboard sa fie neambiguu si sa poata servi
direct ca parametru de request fara riscuri de traversare de cale.
+DOUA DIRECTOARE, un singur spatiu de nume:
+ documents/ oglinda Google Drive. `rclone sync` STERGE de aici tot ce nu
+ exista in Drive — nu pune nimic de mana.
+ documents-local/ documente care nu vin din Drive (dictionare proprii, ce se
+ adauga din dashboard). Sincronizarea nu se atinge de ele.
+Cand acelasi nume apare in ambele, castiga cel din Drive: e sursa comuna a
+echipei, iar copia locala ar putea fi o versiune veche uitata acolo.
+
Cand acelasi document exista in mai multe formate (`d406_saft_knowledge.xml` si
`d406_saft_knowledge.md`), la indexare intra UNUL SINGUR, cel cu formatul cel mai
bogat: .xml > .md > .txt. Sursele .xml sunt structurate pe probleme
@@ -38,13 +46,32 @@ def validate_name(name: str) -> str:
return name
+def _all_files() -> list[pathlib.Path]:
+ """Fisierele din ambele directoare. La acelasi nume, cel din Drive castiga."""
+ vazute: dict[str, pathlib.Path] = {}
+ for d in (config.DOCS_LOCAL_DIR, config.DOCS_DIR): # Drive ultimul = suprascrie
+ if not d.is_dir():
+ continue
+ for f in sorted(d.glob("*")):
+ if f.is_file() and f.suffix in DOC_EXTENSIONS:
+ vazute[f.name] = f
+ return [vazute[n] for n in sorted(vazute)]
+
+
+def path_for(name: str) -> pathlib.Path | None:
+ """Unde e documentul, in oricare din cele doua directoare."""
+ name = validate_name(name)
+ for d in (config.DOCS_DIR, config.DOCS_LOCAL_DIR):
+ p = d / name
+ if p.is_file():
+ return p
+ return None
+
+
def list_documents() -> list[dict]:
"""Tot ce e in depozit. `shadowed_by` spune de ce un fisier nu ajunge in index."""
config.DOCS_DIR.mkdir(parents=True, exist_ok=True)
- files = [
- f for f in sorted(config.DOCS_DIR.glob("*"))
- if f.is_file() and f.suffix in DOC_EXTENSIONS
- ]
+ files = _all_files()
# castigatorul per nume de baza: formatul cel mai bogat
winners: dict[str, str] = {}
for f in files:
@@ -60,6 +87,7 @@ def list_documents() -> list[dict]:
"name": f.name,
"size": st.st_size,
"mtime": st.st_mtime,
+ "local": f.parent == config.DOCS_LOCAL_DIR,
"shadowed_by": None if winner == f.name else winner,
})
return out
@@ -71,20 +99,27 @@ def documents_for_index() -> list[dict]:
def read_document(name: str) -> str:
- name = validate_name(name)
- return (config.DOCS_DIR / name).read_text(encoding="utf-8")
+ path = path_for(name)
+ if path is None:
+ raise FileNotFoundError(name)
+ return path.read_text(encoding="utf-8")
def write_document(name: str, content: str) -> None:
+ """Scrie in documents-local/, NU in oglinda Drive.
+
+ Un document scris in `documents/` dispare la urmatorul `rclone sync` — asta
+ se intampla pana acum cu tot ce se adauga din dashboard.
+ """
name = validate_name(name)
- config.DOCS_DIR.mkdir(parents=True, exist_ok=True)
- (config.DOCS_DIR / name).write_text(content, encoding="utf-8")
+ config.DOCS_LOCAL_DIR.mkdir(parents=True, exist_ok=True)
+ (config.DOCS_LOCAL_DIR / name).write_text(content, encoding="utf-8")
def delete_document(name: str) -> bool:
- name = validate_name(name)
- path = config.DOCS_DIR / name
- if not path.exists():
+ """Sterge documentul de unde se afla. Cel din Drive revine la sincronizare."""
+ path = path_for(name)
+ if path is None:
return False
path.unlink()
return True
diff --git a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/sync.py b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/sync.py
index 3fd57d6..300d037 100644
--- a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/sync.py
+++ b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/sync.py
@@ -32,10 +32,11 @@ def _fingerprint() -> str:
doar cand s-a schimbat efectiv ceva, nu la fiecare tur de sincronizare."""
config.DOCS_DIR.mkdir(parents=True, exist_ok=True)
h = hashlib.sha256()
- for f in sorted(config.DOCS_DIR.glob("*")):
- if f.is_file() and f.suffix in store.DOC_EXTENSIONS:
- st = f.stat()
- h.update(f"{f.name}:{st.st_mtime_ns}:{st.st_size}\n".encode())
+ # amandoua directoarele: si un document local adaugat din dashboard trebuie sa
+ # declanseze reindexarea, nu doar ce vine din Drive
+ for f in store._all_files():
+ st = f.stat()
+ h.update(f"{f.name}:{st.st_mtime_ns}:{st.st_size}\n".encode())
return h.hexdigest()
diff --git a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/tests/test_store_si_chunking.py b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/tests/test_store_si_chunking.py
index d463659..ec17477 100644
--- a/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/tests/test_store_si_chunking.py
+++ b/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/tests/test_store_si_chunking.py
@@ -237,3 +237,57 @@ def test_textul_modificat_se_reembedeaza(monkeypatch, write):
assert calculate == ["varianta modificata"]
assert rezultat["embeddings_refolosite"] == 0
+
+
+# --- documente locale, in afara oglinzii Drive -------------------------------
+
+def _scrie_local(name: str, text: str):
+ import config
+ config.DOCS_LOCAL_DIR.mkdir(parents=True, exist_ok=True)
+ p = config.DOCS_LOCAL_DIR / name
+ p.write_text(text, encoding="utf-8")
+ return p
+
+
+def test_documentele_locale_intra_in_index(write):
+ import store
+ write("din-drive.md", "vine din Drive")
+ _scrie_local("al-meu.md", "adaugat local")
+
+ nume = {d["name"] for d in store.documents_for_index()}
+ assert nume == {"din-drive.md", "al-meu.md"}
+
+
+def test_se_vede_care_e_local(write):
+ import store
+ write("din-drive.md", "x")
+ _scrie_local("al-meu.md", "y")
+ dupa_nume = {d["name"]: d for d in store.list_documents()}
+ assert dupa_nume["al-meu.md"]["local"] is True
+ assert dupa_nume["din-drive.md"]["local"] is False
+
+
+def test_la_acelasi_nume_castiga_versiunea_din_drive(write):
+ import store
+ _scrie_local("acelasi.md", "copie locala veche")
+ write("acelasi.md", "versiunea din Drive")
+ assert store.read_document("acelasi.md") == "versiunea din Drive"
+ assert len(store.list_documents()) == 1
+
+
+def test_scrierea_din_dashboard_merge_in_directorul_local():
+ # Pana acum ajungea in oglinda Drive si disparea la urmatorul rclone sync.
+ import config
+ import store
+ store.write_document("nota.md", "ceva de tinut minte")
+ assert (config.DOCS_LOCAL_DIR / "nota.md").exists()
+ assert not (config.DOCS_DIR / "nota.md").exists()
+
+
+def test_stergerea_gaseste_documentul_in_ambele_directoare(write):
+ import store
+ _scrie_local("local.md", "x")
+ write("drive.md", "y")
+ assert store.delete_document("local.md") is True
+ assert store.delete_document("drive.md") is True
+ assert store.delete_document("inexistent.md") is False