Google Drive Sync for Session-Memoria

Skill ID: gdrive-sync-memoria Version: 1.0.0 Auto-load: No (manual trigger or proactive reminder) Dependencies: session-memoria

Purpose

Integrates Google Drive with session-memoria to automatically import summaries and documents from ChatLLM Teams (ABACUS.AI) into Jimmy's knowledge base.

Hybrid strategy: Use ChatLLM Teams for heavy processing (token-intensive tasks), then import the resulting .md summaries into session-memoria via Google Drive.

Workflow:

•Jimmy creates summaries in ChatLLM Teams (English or Portuguese)
•Saves .md files to Google Drive folder _tobe_registered
•Xavier syncs files → processes → creates session-memoria entries
•Renames files with _pushed_session_memoria suffix
•Moves processed files to _registered_claude_session_memoria
•Git commit + push changes

Configuration

File: config/drive_folders.json

json

{
  "input_folder": "_tobe_registered",
  "input_folder_id": "1Lmf51XNKbnzdBYlR0VgsWhP6YjgF3si7",
  "processed_folder": "_registered_claude_session_memoria",
  "processed_folder_id": "1fF4kTEUTEbG-Jf9tGdV5Ur7bnnW1bmLN",
  "rclone_remote": "gdrive-jimmy",
  "preserve_original_language": true,
  "auto_categorize": true,
  "max_tags": 5,
  "summary_max_length": 120
}

Required setup (one-time):

•rclone installed and configured
•Remote named gdrive-jimmy with full Google Drive access
•Folders _tobe_registered and _registered_claude_session_memoria created

Triggers (Portuguese)

Manual

•/gdrive-sync
•"Xavier, sincroniza o Google Drive"
•"X, processa arquivos do ChatLLM"
•"Importa os resumos do Google Drive"

Proactive Reminder

Track last sync in session-memoria/knowledge/metadata.json:

json

"gdrive_sync": {
  "enabled": true,
  "last_sync": "2026-02-11 22:00:00",
  "total_synced": 15,
  "last_count": 3
}

Reminder logic:

•
If last_sync is null OR > 3 days ago:
- •"Jimmy, faz 3 dias sem sincronizar Google Drive. Quer processar agora?"
- •Only remind once per session (don't spam)

Workflow (8 Steps)

Pre-flight Validations

Before starting sync, verify:

•
rclone configured?
bash
```
rclone version
rclone lsd gdrive-jimmy:
```
- •If fails: "❌ rclone não configurado. Execute: rclone config"
•
Git repo clean?
bash
```
cd claude-intelligence-hub
git status
```
- •If dirty: Alert Jimmy (uncommitted changes exist)
- •Proceed anyway (sync adds new files)
•
metadata.json valid?
bash
```
cat session-memoria/knowledge/metadata.json | jq .
```
- •If invalid: "❌ metadata.json corrupted. Cannot proceed."

If all validations pass → proceed with sync.

Step 1: List Files in Input Folder

bash

rclone lsf gdrive-jimmy:_tobe_registered/ --files-only --include "*.md"

Output: List of .md filenames (one per line)

If empty:

•"✅ Nenhum arquivo pendente no Google Drive."
•Update gdrive_sync.last_sync in metadata.json
•Exit gracefully

If files found:

•Log count: "📥 Encontrados X arquivos para processar"
•Proceed to Step 2 for each file

Step 2: Download File

For each file from Step 1:

bash

rclone copy "gdrive-jimmy:_tobe_registered/[FILENAME]" ./gdrive-sync-memoria/temp/ -v

Validations:

•File exists in temp/?
•File size > 0 bytes?
•File is valid UTF-8?

If validation fails:

•Log error: logs/sync_history.log
•Skip file, continue next
•Error format: [TIMESTAMP] ERROR: [FILENAME] - [REASON]

Step 3: Parse File Content

Read file:

bash

cat ./gdrive-sync-memoria/temp/[FILENAME]

Extract metadata:

•
Title:
- •First H1 heading (# Title)
- •If no H1: use filename (without .md)
- •Max 80 chars
•
Content:
- •Full markdown content (preserve original)
- •NO translation (keep English, Portuguese, or any language)
- •Strip excessive whitespace (max 2 consecutive newlines)
•
Category (auto-infer):
- •
  Scan content for keywords:
  - •"Power BI", "DAX", "Power Query", "M language" → Power BI
  - •"Python", "script", "pandas", "numpy" → Python
  - •"project", "team", "gestão", "management" → Gestão
  - •"API", "endpoint", "REST", "GraphQL" → API Development
  - •"SQL", "database", "query" → Database
  - •Default → Other
- •If multiple matches: use first match
- •Jimmy can recategorize later via session-memoria
•
Tags (auto-extract):
- •Extract up to 5 tags from content
- •Prioritize: section headers, bold terms, code blocks
- •Format: kebab-case
- •Examples: power-bi, dax-measures, python-automation
- •If <5 tags: supplement with category-based defaults
•
Summary:
- •Extract from first paragraph OR
- •Generate concise summary from content
- •Preserve original language (NO translation)
- •Max 120 chars
- •Focus: what problem does this solve? what knowledge is captured?

Step 4: Create Session-Memoria Entry

Follow existing Save workflow (session-memoria SKILL.md, lines 141-243)

Entry ID generation:

code

YYYY-MM-DD-NNN
Example: 2026-02-11-001

Frontmatter template:

yaml

---
id: 2026-02-11-001
title: [Extracted title]
summary: [Generated summary - original language]
category: [Auto-inferred category]
tags:
  - [tag-1]
  - [tag-2]
  - [tag-3]
date: 2026-02-11
source: chatllm-teams-gdrive
original_filename: [FILENAME]
language: [auto-detected: en, pt, etc.]
---

[Full original content]

File path:

code

session-memoria/knowledge/entries/2026/02/2026-02-11-001.md

Index updates (same as normal Save):

•

by-date index:

code

session-memoria/knowledge/indexes/by-date/2026-02.md

Append:

markdown

## 2026-02-11
- [2026-02-11-001](../../entries/2026/02/2026-02-11-001.md) - [summary] `#tag-1` `#tag-2`

•

by-category index:

code

session-memoria/knowledge/indexes/by-category/[category].md

Append:

markdown

- [2026-02-11-001](../../entries/2026/02/2026-02-11-001.md) - [summary]

•

by-tag indexes: For each tag, update:

code

session-memoria/knowledge/indexes/by-tag/[tag].md

Append:

markdown

- [2026-02-11-001](../../entries/2026/02/2026-02-11-001.md) - [summary]

•

metadata.json:

json

{
  "total_entries": [increment by 1],
  "last_updated": "2026-02-11T22:30:00Z",
  "last_entry_id": "2026-02-11-001",
  "gdrive_sync": {
    "last_sync": "2026-02-11T22:30:00Z",
    "total_synced": [increment by 1],
    "last_count": [files processed this run]
  }
}

Step 5: Rename File in Google Drive

bash

original="[FILENAME].md"
renamed="${original%.md}_pushed_session_memoria.md"

rclone moveto \
  "gdrive-jimmy:_tobe_registered/$original" \
  "gdrive-jimmy:_tobe_registered/$renamed" \
  --drive-server-side-across-configs

Why rename before move?

•Atomic operation visibility
•If Step 6 fails, file is marked as processed (won't re-process)
•Easy to identify processed files in input folder

Validation:

•File renamed successfully?
•Original file no longer exists in input folder?

Step 6: Move to Processed Folder

bash

renamed="${original%.md}_pushed_session_memoria.md"

rclone moveto \
  "gdrive-jimmy:_tobe_registered/$renamed" \
  "gdrive-jimmy:_registered_claude_session_memoria/$renamed" \
  --drive-server-side-across-configs

Server-side move benefits:

•Fast (no download/upload)
•Preserves metadata
•Atomic operation

Validation:

•File exists in processed folder?
•File removed from input folder?

Step 7: Git Commit + Push

After processing ALL files in batch:

bash

cd claude-intelligence-hub

git add session-memoria/knowledge/

git commit -m "feat(session-memoria): sync from Google Drive - [COUNT] files

Processed files:
- [FILENAME_1]
- [FILENAME_2]
- [...]

Source: ChatLLM Teams via gdrive-sync-memoria
Categories: [list unique categories]
Total entries added: [COUNT]"

git push origin main

Retry logic:

•Retry up to 3 times with exponential backoff (2s, 4s, 8s)
•
If all retries fail:
- •"⚠️ Git push failed. Entries created locally but not synced."
- •Log error to logs/sync_history.log
- •Alert Jimmy (requires manual intervention)

Why batch commit vs. per-file?

•Reduces commit noise
•Atomic sync operation
•Easier rollback if needed

Step 8: Cleanup and Logging

Delete temp files:

bash

rm -f ./gdrive-sync-memoria/temp/*.md

Log success:

code

[2026-02-11 22:30:15] SUCCESS: Synced 3 files
- file1.md → 2026-02-11-001 (Power BI)
- file2.md → 2026-02-11-002 (Python)
- file3.md → 2026-02-11-003 (Other)

Update metadata.json:

json

"gdrive_sync": {
  "last_sync": "2026-02-11T22:30:15Z",
  "total_synced": 18,  // previous: 15
  "last_count": 3,
  "last_error": null
}

User feedback:

code

✅ Sincronização completa!

📥 3 arquivos processados
📝 Entries criadas: 2026-02-11-001, 002, 003
📂 Categorias: Power BI (1), Python (1), Other (1)
🔄 Git push: success
📦 Arquivos movidos para _registered_claude_session_memoria

Próxima sincronização sugerida: 2026-02-14

Error Handling

Network Failures

Scenario: rclone commands timeout or fail

Strategy: Retry with exponential backoff

python

max_retries = 3
backoff = [2, 4, 8]  # seconds

for attempt in range(max_retries):
    result = rclone_command()
    if result.success:
        break
    if attempt < max_retries - 1:
        sleep(backoff[attempt])
    else:
        log_error("Network failure after 3 retries")
        alert_jimmy("❌ Falha de rede ao acessar Google Drive")
        abort_sync()

Logged in: logs/sync_history.log

Malformed Files

Scenarios:

•Empty file (0 bytes)
•Corrupted UTF-8
•No markdown content
•Title extraction fails

Strategy: Skip file, log error, continue next

code

[2026-02-11 22:15:30] ERROR: corrupt.md - File is empty (0 bytes)
[2026-02-11 22:15:35] ERROR: bad_encoding.md - Invalid UTF-8 encoding

User alert (end of sync):

code

⚠️ 2 arquivos ignorados (ver logs/sync_history.log):
- corrupt.md (empty file)
- bad_encoding.md (encoding error)

DO NOT:

•Stop entire sync due to one bad file
•Attempt to "fix" corrupted files automatically
•Move bad files to processed folder

Leave bad files in input folder for Jimmy to manually review.

Git Push Failures

Scenario: Network issues, merge conflicts, auth failures

Strategy:

•Retry up to 3 times (2s, 4s, 8s backoff)
•
If all fail:
- •Log error
- •Alert Jimmy
- •DO NOT delete local changes

Alert message:

code

⚠️ Git push failed após 3 tentativas

✅ Entries criadas localmente:
- session-memoria/knowledge/entries/2026/02/2026-02-11-001.md
- session-memoria/knowledge/entries/2026/02/2026-02-11-002.md

❌ Sincronização remota pendente

Ação necessária:
1. Verificar conectividade
2. Executar manualmente: cd claude-intelligence-hub && git push

Recovery path:

•Jimmy can manually run git push later
•Local entries are safe
•Files already moved in Google Drive (won't re-process)

Duplicate Files

Scenario: File with same name already processed

Detection:

•Check if [FILENAME]_pushed_session_memoria.md exists in input folder
•Check if file exists in processed folder

Strategy:

•Skip file
•Log warning: "⚠️ [FILENAME] already processed (duplicate)"
•Continue next file

Why not auto-delete?

•Jimmy may have intentionally uploaded updated version
•Manual review safer than auto-deletion

Validation Failures

Pre-flight failures:

Check	Failure Action
rclone not configured	Abort sync, show setup instructions
metadata.json corrupted	Abort sync, alert Jimmy
Git repo not found	Abort sync, show error

Per-file failures:

Check	Failure Action
File empty	Skip, log error
Invalid UTF-8	Skip, log error
Summary > 120 chars	Truncate to 120
Tags > 5	Keep first 5
Title > 80 chars	Truncate to 80

Logging

File: logs/sync_history.log

Format:

code

[YYYY-MM-DD HH:MM:SS] [LEVEL] [FILENAME] - Message

Examples:

code

[2026-02-11 22:15:00] INFO START: Sync initiated
[2026-02-11 22:15:02] INFO FOUND: 3 files in _tobe_registered
[2026-02-11 22:15:05] SUCCESS PROCESS: file1.md → 2026-02-11-001 (Power BI)
[2026-02-11 22:15:10] ERROR SKIP: corrupt.md - File is empty (0 bytes)
[2026-02-11 22:15:15] SUCCESS PROCESS: file2.md → 2026-02-11-002 (Python)
[2026-02-11 22:15:20] SUCCESS COMMIT: Git commit successful
[2026-02-11 22:15:25] SUCCESS PUSH: Git push successful
[2026-02-11 22:15:30] INFO COMPLETE: Synced 2/3 files, 1 skipped

Rotation:

•Keep last 1000 lines
•Rotate when > 100KB
•Archive old logs (sync_history_YYYYMMDD.log)

NOT version controlled (git-ignored)

Proactive Monitoring

Trigger reminder when:

•gdrive_sync.last_sync is null (never synced)
•gdrive_sync.last_sync > 3 days ago
•Not already reminded this session

Reminder message:

code

Jimmy, faz 3 dias sem sincronizar Google Drive.

Última sincronização: 2026-02-08 (3 dias atrás)
Total sincronizado: 15 arquivos

Quer processar novos arquivos agora?

User responses:

•"Sim" / "Processa" / "Vai" → Execute sync
•"Não" / "Depois" → Skip (don't remind again this session)
•"Desabilita lembretes" → Set gdrive_sync.enabled = false in metadata.json

Integration with Session-Memoria

Entry metadata additions:

Standard session-memoria frontmatter + these fields:

yaml

source: chatllm-teams-gdrive
original_filename: power-bi-summary.md
language: en  # or pt, auto-detected
sync_date: 2026-02-11T22:15:05Z

Why track original_filename?

•Debugging (trace back to Google Drive file)
•Avoid duplicate processing
•Audit trail

Why track language?

•Helps future AI understand content context
•Enables language-specific search/filtering
•Preserves intent (no translation artifacts)

Usage Examples

Example 1: Manual Sync

Jimmy:

code

Xavier, sincroniza o Google Drive

Xavier:

code

🔄 Sincronizando Google Drive...

📥 Encontrados 2 arquivos:
- power-bi-best-practices.md
- python-automation-summary.md

Processando...

✅ Sincronização completa!

📝 Entries criadas:
- 2026-02-11-001: Power BI Best Practices (Power BI)
- 2026-02-11-002: Python Automation Summary (Python)

🔄 Git push: success
📦 Arquivos movidos para _registered_claude_session_memoria

Example 2: Proactive Reminder

Xavier (proactive):

code

Jimmy, faz 3 dias sem sincronizar Google Drive.

Última sincronização: 2026-02-08
Total sincronizado: 15 arquivos

Quer processar novos arquivos agora?

Jimmy:

code

Vai

Xavier:

code

🔄 Sincronizando...

✅ Nenhum arquivo pendente no Google Drive.
Tudo atualizado!

Example 3: Error Handling

Xavier:

code

🔄 Sincronizando Google Drive...

📥 Encontrados 3 arquivos

⚠️ Erros encontrados:
- empty-file.md: Arquivo vazio (0 bytes) - ignorado
- corrupt.md: Encoding inválido - ignorado

✅ 1 arquivo processado:
- 2026-02-11-001: DAX Measures Guide (Power BI)

📋 Ver detalhes: gdrive-sync-memoria/logs/sync_history.log

Setup Instructions (One-Time)

1. Install rclone

Windows (Scoop):

bash

scoop install rclone

Or manual download: https://rclone.org/downloads/

Verify:

bash

rclone version

2. Configure Google Drive Remote

bash

rclone config

Interactive prompts:

code

n) New remote
name> gdrive-jimmy
Storage> drive  (Google Drive)
client_id> [Enter] (use default)
client_secret> [Enter] (use default)
scope> 1  (Full access)
root_folder_id> [Enter] (entire drive)
service_account_file> [Enter] (none)
Advanced config? n
Use auto config? y  (Opens browser for OAuth)

[Authenticate in browser]

Configure as team drive? n

Verify:

bash

rclone lsd gdrive-jimmy:

Should list all folders in Google Drive.

3. Verify Folder Access

bash

rclone lsf gdrive-jimmy:_tobe_registered/
rclone lsf gdrive-jimmy:_registered_claude_session_memoria/

If folders don't exist:

bash

rclone mkdir gdrive-jimmy:_tobe_registered
rclone mkdir gdrive-jimmy:_registered_claude_session_memoria

4. Update .gitignore

Add to claude-intelligence-hub/.gitignore:

code

gdrive-sync-memoria/logs/
gdrive-sync-memoria/temp/

Commit:

bash

git add .gitignore
git commit -m "chore: ignore gdrive-sync-memoria temp files"

5. First Sync Test

Create test file:

bash

echo "# Test Summary from ChatLLM" > test.md
rclone copy test.md gdrive-jimmy:_tobe_registered/

Run sync:

code

/gdrive-sync

Expected:

•✅ File downloaded to temp/
•✅ Entry created in session-memoria
•✅ Indexes updated
•✅ Git commit + push
•✅ File moved to _registered_claude_session_memoria

Security Considerations

rclone authentication:

•Uses OAuth tokens (stored in ~/.config/rclone/rclone.conf)
•Token refresh automatic
•Revoke access: Google Account settings → Apps

Sensitive data:

•NO passwords/API keys in .md files
•Review files before uploading to Google Drive
•gdrive-sync-memoria does NOT filter secrets (Jimmy's responsibility)

Git commits:

•All entries are public (if repo is public)
•Review content before pushing
•Use private repo for sensitive knowledge

Troubleshooting

"rclone: command not found"

Solution:

bash

# Install rclone
scoop install rclone

# Or add to PATH manually

"Failed to create file system: didn't find section in config file"

Solution:

bash

# Remote not configured
rclone config

# Or check spelling: gdrive-jimmy (exact match)

"403 Forbidden" from Google Drive

Solution:

bash

# Re-authenticate
rclone config reconnect gdrive-jimmy:

# Or delete and re-create remote
rclone config delete gdrive-jimmy
rclone config  # Create new

"metadata.json: Permission denied"

Solution:

bash

# File locked by another process
# Close Claude Code, VS Code, or text editors
# Retry sync

Files not appearing in session-memoria

Debug checklist:

•Check logs/sync_history.log for errors
•Verify entry created: session-memoria/knowledge/entries/YYYY/MM/
•Verify indexes updated: session-memoria/knowledge/indexes/
•Check git status: git log -1 --oneline
•Verify metadata.json updated

Performance

Typical sync times:

Files	Duration
1 file	~10 seconds
5 files	~30 seconds
10 files	~60 seconds
20 files	~2 minutes

Bottlenecks:

•Network latency (Google Drive API)
•Git push (remote sync)
•Entry parsing (minimal)

Optimization:

•Server-side move (no download/upload)
•Batch git commits
•Parallel processing (future enhancement)

Future Enhancements (v2.0)

Not in scope for v1.0:

• Parallel file processing
• Web UI for monitoring
• Auto-delete old files from processed folder
• Conflict resolution (duplicate titles)
• Multi-language translation (preserve original + add PT/EN)
• Image embedding support
• PDF conversion support
• Automated categorization via AI (Claude API)

Request features: Jimmy can suggest enhancements based on usage patterns.

Version History

See CHANGELOG.md for detailed version history.

Current version: 1.0.0 (2026-02-11)