From 15c3ef1c5e8ca7b4b09db84fce2687f60a6a4762 Mon Sep 17 00:00:00 2001 From: Canyon Robins Date: Wed, 14 May 2025 14:45:28 -0700 Subject: [PATCH] [Condense] Add docs for intelligent condense experimental feature --- docs/features/experimental/experimental-features.md | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/docs/features/experimental/experimental-features.md b/docs/features/experimental/experimental-features.md index a6a59a1f..5f7a7281 100644 --- a/docs/features/experimental/experimental-features.md +++ b/docs/features/experimental/experimental-features.md @@ -18,6 +18,12 @@ To enable or disable experimental features: The following experimental features are currently available: +### Intelligently condense the context window + +If enabled, Roo uses an LLM to summarize the past conversation when the context window is almost full. This summary attempts to dramatically shrink the number of tokens in the context while preserving useful information. If disabled, Roo drops past messages using a sliding window when the context window is full. + +**Disclaimer**: Currently the costs associated with summarizing the conversation are not included in the cost displayed in the UI. + ### Power Steering When enabled, Roo will remind the model about the details of its current mode definition more frequently. This will lead to stronger adherence to role definitions and custom instructions, but will use more tokens per message.