Feature Request: Improve context management

[nissa-seru]

IIUC, the current logic garbage-collects the first-half of the conversation (except for the original task content) upon hitting ~80% model context length. A few optimizations to consider:

• Consider either:
• Setting a cache checkpoint after the task + environment details + contents of @mentioned files; or
• Use the opportunity from the cache being broken to update the environment details, and contents of @mentioned files.

I would recommend doing one or the other as a "free" optimization, especially as the original task + context can be quite bulky in terms of tokens.

Having multiple versions of the same file in context:

• Easily happens if a file is read multiple times over the course of work (especially in the event if an edit-miss requiring file re-read)
• Consumes a massive amount of tokens
• Greatly decreases output quality because of the relative textual similarity of the file versions simultaneously coexisting in memory.

When context is being garbage-collected, I would recommend:

1. Cleanup of any duplicate file contents in context (favoring the newest one); and
2. Prioritizing retention of user text -> model text -> read-file results -> write-file results, to limit the amount of abrupt amnesia that impacts the model at the garbage collection point.

[vijaykrishna00]

https://github.com/GreatScottyMac/roo-code-memory-bank

i found this and seem to be helpful to me
we can implement an improved version of this by default

[dmanresa-saes]

I'm wondering if anyone knows whether the Cline Memory Bank (remembering that Roo Code is a fork of Cline) is compatible with Roo Code. The Cline Memory Bank looks simpler or at least more straightforward compared to the memory bank available at https://github.com/GreatScottyMac/roo-code-memory-bank . This one might work better, but I haven't had the time to read through the prompts to fully understand them.

[vitalii4reva]

Hi everyone!

Looking at this discussion, I see we're all facing similar challenges with context management. The current memory bank solutions seem focused on storing predefined context, but I think we need something more dynamic.

What if, instead of just managing what we already have in context, the AI could proactively discover what context it needs?

My proposal builds on these ideas but takes a different approach:

1. Self-guided discovery - When Claude detects it's missing context (like a function it references but doesn't have), it automatically searches the codebase to find it

2. Smart prioritization - Beyond just cleaning up duplicates, it would rank files by relevance to the current query and adaptively manage the context window

3. Integrated context management - Rather than requiring manual @mentions of files, it would intelligently determine what should be in context

I've been experimenting with LangChain for implementing this kind of system, and it provides good tools for memory management that could address the exact issues @nissa-seru mentioned about garbage collection priorities and duplicates.

This would go beyond what the current memory banks provide by being adaptive rather than static.

Would this direction be interesting to explore? I'd be happy to elaborate more on the technical approach or even work on a proof of concept if there's interest!

[ebynnerlys]

🌐 Versión en Español

📂 Solicitud: Mejoras en la Gestión de Archivos Duplicados y Contexto en RooCode

Descripción:

1. 🗂️ Detección automática de archivos duplicados:

   • Identificar y eliminar automáticamente versiones repetidas de un mismo archivo, conservando solo la más reciente.
   • Notificar al usuario cuando se detecten duplicados para mayor transparencia.

2. 📊 Panel de contexto mejorado:

   • Mostrar una lista clara y organizada de los archivos en contexto, resaltando aquellos que están activos y los que han sido descartados por el garbage collection.
   • Permitir al usuario excluir manualmente archivos específicos del contexto, incluso si no son duplicados.

3. 🧠 Priorización inteligente de retención:

   • Implementar un sistema que priorice la retención de:
     • Texto del usuario (instrucciones).
     • Respuestas del modelo.
     • Resultados de archivos leídos (@mentioned).
   • Esto reduciría la "amnesia abrupta" durante el garbage collection.

Beneficios:

• 🎯 Optimización de tokens: Evitar el desperdicio de tokens por archivos duplicados.
• ⏱ Mayor eficiencia: Mejorar la calidad de las respuestas al mantener solo la información relevante.
• 🧠 Control y transparencia: Permitir al usuario gestionar activamente el contexto.

---

🌐 English Version

📂 Request: Improvements in Duplicate File Management and Context Handling in RooCode

Description:

1. 🗂️ Automatic detection of duplicate files:

   • Identify and automatically remove repeated versions of the same file, keeping only the most recent one.
   • Notify the user when duplicates are detected for greater transparency.

2. 📊 Enhanced context panel:

   • Display a clear and organized list of files in context, highlighting active files and those discarded by garbage collection.
   • Allow the user to manually exclude specific files from the context, even if they are not duplicates.

3. 🧠 Smart retention prioritization:

   • Implement a system that prioritizes retaining:
     • User text (instructions).
     • Model responses.
     • Results from read files (@mentioned).
   • This would reduce "abrupt amnesia" during garbage collection.

Benefits:

• 🎯 Token optimization: Avoid wasting tokens on duplicate files.
• ⏱ Greater efficiency: Improve response quality by keeping only relevant information.
• 🧠 Control and transparency: Allow users to actively manage the context.