BundleMaker v2.2.0

LLM-friendly project bundler with patch apply, verify, and restore workflow support.

BundleMaker is a Python CLI tool for packaging an entire source project into a single bundle.txt, sending it to an LLM for review or modification, then safely applying the modified result back to the real project with verification and rollback support.

---

Why BundleMaker?

When using LLMs on multi-file projects, context often breaks down:

• only part of the code is considered
• project intent and implementation become separated
• modified results are hard to patch back safely
• rollback is painful when something goes wrong

BundleMaker was built to reduce that friction.

It supports a full workflow:

Project → Bundle → LLM modification → Patch apply → Verify → Restore

---

Features

Bundle creation

Creates a single bundle.txt from multiple project files.

Supported modes:

• AUTO — automatically reads files and builds a bundle
• PASTE — manually paste file contents
• HYBRID — auto-read first, manual fallback if needed

Patch apply

Compares the original bundle and the modified bundle, then applies the changes back to the real project.

Strategies:

• SAFE — small changes + new files only, no deletes
• FULL — apply all changes including deletes
• DRY RUN — report only, no actual writes

Verification

Performs real post-apply verification.

• VERIFY-A — compares modified bundle against actual disk files
• VERIFY-B — rebuilds a fresh bundle from disk and compares again

Restore / rollback

Supports rollback using:

• manifest records
• file backups
• deleted file snapshots
• restore verification
• created file quarantine

Config wizard

Interactive generation of .bundlemaker.json for:

• allowed extensions
• excluded directories

---

Typical workflow

1. Create a bundle

python bundlemaker.py --auto

This generates:

bundles/bundle.txt

2. Send the bundle to your LLM

A typical prompt example:

1.bundle.txt에는 모든 코드가 들어있고 master문서(EX.개발 명세서) 에는 이 프로젝트의 기획이 담겨있다. 이 모든걸 종합적으로 꼼꼼히 확인하여 지금 코드에서 나타나는 문제점과 버그를 알려줘라

Then save the modified result as:

bundle_modified.txt

3. Apply the patch

python bundlemaker.py --patch

4. Verify the result

BundleMaker automatically runs verification after patching.

5. Restore if needed

python bundlemaker.py --restore

---

Main menu

[1] STEP 1 — Create bundle from project
[2] STEP 3 — Apply patch from modified bundle (+VERIFY)
[c] Configure — create/overwrite .bundlemaker.json
[r] Restore — rollback from patch manifest backups
[q] Quit

---

CLI options

Configure

python bundlemaker.py --config
python bundlemaker.py --configure

Bundle creation

python bundlemaker.py --auto
python bundlemaker.py --paste
python bundlemaker.py --hybrid

Patch apply

python bundlemaker.py --patch
python bundlemaker.py --apply-patch

Restore

python bundlemaker.py --restore
python bundlemaker.py --rollback

---

Directory layout

Internal artifacts are stored under the project root:

<PROJECT_ROOT>/
└─ .bundlemaker/
   ├─ state/
   ├─ manifests/
   ├─ reports/
   ├─ backups/
   ├─ locks/
   └─ quarantine/

Generated bundles are stored under the current working directory:

<CWD>/
└─ bundles/
   ├─ bundle.txt
   └─ bundle_applied_<timestamp>.txt

These paths are automatically excluded from scanning:

• .bundlemaker
• bundles

---

Default supported extensions

py, sql,
html, css, js,
c, h,
cpp, hpp, cc, hh,
cs

---

Default excluded directories

.git, .svn, .hg
__pycache__, .pytest_cache
node_modules
venv, .venv
dist, build
.idea, .vscode
.bundlemaker
bundles

---

Bundle format

Generated bundles follow this structure:

=== BUNDLE GENERATED: 2026-03-09T12:34:56 ===
=== REL_ROOT: C:\project ===

=== FILE: src/main.py ===
print("hello")
=== END FILE: src/main.py ===

---

Safety features

BundleMaker includes:

• atomic file writes
• root boundary checks
• path traversal prevention
• manifest-based patch tracking
• backup handling
• duplicate STEP 3 lock
• disk verification after patch
• restore verification after rollback
• quarantine support for created files

---

Auto-read safeguards

Auto-read may skip:

• files larger than 5MB
• binary files
• sensitive names such as .env, id_rsa
• sensitive extensions such as .pem, .key, .pfx, .crt

---

Recommended usage order

1. Configure rules
2. Create bundle
3. Send bundle to LLM
4. Save modified result as bundle_modified.txt
5. Run patch
6. Check verify reports
7. Restore if necessary

For important projects, start with:

• DRY RUN
• then SAFE
• use FULL only when needed

---

Current status

BundleMaker is currently a personal experimental tool focused on text-based source projects and LLM-assisted workflows.

It is practical for personal use and internal tooling, but still evolving.

---

Best use cases

• multi-file LLM code review
• whole-project context packaging
• safer LLM-assisted patch workflows
• educational / experimental coding
• internal developer tooling
• research prototypes

---

Summary

BundleMaker v2.2.0 helps turn multi-file codebases into a single LLM-friendly bundle, then safely apply, verify, and restore changes through a structured workflow.