Fimeg/letta-server
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Add modes self and self_sliding_window for prompt caching (#9372)

Browse Source 
* add self compaction method with proper caching (pass in tools, don't refresh sys prompt beforehand) + sliding fallback

* updated prompts for self compaction

* add tests for self, self_sliding_window modes and w/o refresh messages before compaction

* add cache logging to summarization

* better handling to prevent agent from continuing convo on self modes

* if mode changes via summarize endpoint, will use default prompt for the new mode

---------

Co-authored-by: Amy Guan <amy@letta.com>
This commit is contained in:
amysguan
2026-02-24 10:15:36 -08:00
committed by Caren Thomas
parent 47d55362a4
commit 47b0c87ebe
15 changed files with 1065 additions and 223 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
fern/openapi.json
View File

@@ -31151,7 +31151,7 @@
}
],
"title": "Model",
"description": "Model handle to use for summarization (format: provider/model-name). If None, uses lightweight provider-specific defaults."
"description": "Model handle to use for sliding_window/all summarization (format: provider/model-name). If None, uses lightweight provider-specific defaults."
},
"model_settings": {
"anyOf": [
@@ -31256,7 +31256,12 @@
},
"mode": {
"type": "string",
"enum": ["all", "sliding_window", "self"],
"enum": [
"all",
"sliding_window",
"self_compact_all",
"self_compact_sliding_window"
],
"title": "Mode",
"description": "The type of summarization technique use.",
"default": "sliding_window"
@@ -31264,12 +31269,12 @@
"sliding_window_percentage": {
"type": "number",
"title": "Sliding Window Percentage",
"description": "The percentage of the context window to keep post-summarization (only used in sliding window mode)."
"description": "The percentage of the context window to keep post-summarization (only used in sliding window modes)."
}
},
"type": "object",
"title": "CompactionSettings",
"description": "Configuration for conversation compaction / summarization.\n\n``model`` is the only required user-facing field – it specifies the summarizer\nmodel handle (e.g. ``\"openai/gpt-4o-mini\"``). Per-model settings (temperature,\nmax tokens, etc.) are derived from the default configuration for that handle."
"description": "Configuration for conversation compaction / summarization.\n\nPer-model settings (temperature,\nmax tokens, etc.) are derived from the default configuration for that handle."
},
"CompactionSettings-Output": {
"properties": {
@@ -31283,7 +31288,7 @@
}
],
"title": "Model",
"description": "Model handle to use for summarization (format: provider/model-name). If None, uses lightweight provider-specific defaults."
"description": "Model handle to use for sliding_window/all summarization (format: provider/model-name). If None, uses lightweight provider-specific defaults."
},
"model_settings": {
"anyOf": [
@@ -31388,7 +31393,12 @@
},
"mode": {
"type": "string",
"enum": ["all", "sliding_window", "self"],
"enum": [
"all",
"sliding_window",
"self_compact_all",
"self_compact_sliding_window"
],
"title": "Mode",
"description": "The type of summarization technique use.",
"default": "sliding_window"
@@ -31396,12 +31406,12 @@
"sliding_window_percentage": {
"type": "number",
"title": "Sliding Window Percentage",
"description": "The percentage of the context window to keep post-summarization (only used in sliding window mode)."
"description": "The percentage of the context window to keep post-summarization (only used in sliding window modes)."
}
},
"type": "object",
"title": "CompactionSettings",
"description": "Configuration for conversation compaction / summarization.\n\n``model`` is the only required user-facing field – it specifies the summarizer\nmodel handle (e.g. ``\"openai/gpt-4o-mini\"``). Per-model settings (temperature,\nmax tokens, etc.) are derived from the default configuration for that handle."
"description": "Configuration for conversation compaction / summarization.\n\nPer-model settings (temperature,\nmax tokens, etc.) are derived from the default configuration for that handle."
},
"CompactionStats": {
"properties": {