AI forgets after long chats? Open-source reimplementation of the Manus-style workflow — “planning-with-files”
Learn how to locate and set up a planning-with-files style open-source workflow, design file-backed memory and prompt retrieval, integrate file I/O with an LLM, and test multi-turn conversations to evaluate persistent context handling.
Image Analysis
Image evidence (frame 1)
Readable text (line-by-line):
1. "planning-with-files:" (original English, retained)
2. "AI" (original uppercase)
3. "失忆?" (Chinese: "Forgetting?")
4. "治它!" (Chinese: "Fix it!")
Visible observations:
- Dark/black background with red brush-splash texture and light speckles.
- High-contrast text colors: most text in white, "治它!" rendered in red.
- Rough, brush/tear-like typography with distressed edges.
- Overall poster/cover style; no UI elements, buttons, or step indicators visible.
Not observable from the image: implementation details, code, interactive steps, or repository links are not present.
SkillPI Creation Analysis
Purpose and theme assessment:
- The image addresses the issue of an LLM or chat AI "forgetting" in long conversations and suggests there is a workflow to "fix" that. The label "planning-with-files:" likely points to a file-based planning or memory-management workflow or a repository name; this should be verified against the original source.
Key image evidence:
- Directly readable strings: "planning-with-files:", "AI", "失忆?", "治它!". The image contains no implementation screenshots or code snippets—it's a promotional/illustrative graphic.
Associated skills and practical steps to attempt a reimplementation (these are recommended, verifiable actions, not claims of reproduction):
1. Locate the project: search for "planning-with-files" in public code hosts or project pages and inspect README/license to confirm availability for reimplementation.
2. Prepare environment: install required runtimes and dependencies as documented (e.g., Python/Node, LLM client libraries) and obtain model API credentials.
3. Design file-driven memory: determine how the workflow encodes context into files (notes, plans, memory files), and prepare sample files to emulate the approach.
4. Prompting and retrieval: craft prompts and retrieval patterns to read/write those files during conversation turns so the model can access prior context.
5. Integration and testing: wire file I/O to model calls, run multi-turn conversation tests, and evaluate whether context persists and is retrieved as intended.
Prerequisites and items to verify:
- Confirm whether "planning-with-files" is an accessible open-source project and identify its implementation language and dependencies.
- Verify any claimed association with "Manus" (author/brand) and check license/usage constraints.
- Because the image contains no code or UI, reproduction requires the original repository or detailed documentation to validate exact mechanisms.
Limitations:
- This interpretation and the suggested steps are based solely on visible text and standard engineering practice. No claim is made that the workflow has been reproduced here; actual reproduction requires the original project materials for verification.