Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

chore(settings): bump default local context window 18432 -> 30208 (#3669)

LearningCircuit15ae51be9f66
diff --git a/src/local_deep_research/defaults/default_settings.json b/src/local_deep_research/defaults/default_settings.json
index 6fc5f7a25c..1754974b19 100644
--- a/src/local_deep_research/defaults/default_settings.json
+++ b/src/local_deep_research/defaults/default_settings.json
@@ -452,7 +452,7 @@
},
"llm.local_context_window_size": {
"category": "llm_parameters",
- "description": "Context window size in tokens for local LLMs (Ollama, LlamaCpp, LM Studio). The default of 18432 tokens is recommended for the langgraph-agent strategy which needs space for tool call history. If you have limited VRAM, reduce this and switch to the source-based strategy instead.",
+ "description": "Context window size in tokens for local LLMs (Ollama, LlamaCpp, LM Studio). The default of 30208 tokens leaves more headroom for the synthesis prompt (sources + previous knowledge + instructions) so it is less likely to be truncated by the model. If you have limited VRAM, reduce this and switch to the source-based strategy instead.",
"editable": true,
"max_value": 131072.0,
"min_value": 512.0,
@@ -461,7 +461,7 @@
"step": 512.0,
"type": "LLM",
"ui_element": "number",
- "value": 18432,
+ "value": 30208,
"visible": true
},
"llm.supports_max_tokens": {
diff --git a/tests/settings/golden_master_settings.json b/tests/settings/golden_master_settings.json
index df44acedcc..aa3dc9703b 100644
--- a/tests/settings/golden_master_settings.json
+++ b/tests/settings/golden_master_settings.json
@@ -1007,7 +1007,7 @@
},
"llm.local_context_window_size": {
"category": "llm_parameters",
- "description": "Context window size in tokens for local LLMs (Ollama, LlamaCpp, LM Studio). The default of 18432 tokens is recommended for the langgraph-agent strategy which needs space for tool call history. If you have limited VRAM, reduce this and switch to the source-based strategy instead.",
+ "description": "Context window size in tokens for local LLMs (Ollama, LlamaCpp, LM Studio). The default of 30208 tokens leaves more headroom for the synthesis prompt (sources + previous knowledge + instructions) so it is less likely to be truncated by the model. If you have limited VRAM, reduce this and switch to the source-based strategy instead.",
"editable": true,
"max_value": 131072.0,
"min_value": 512.0,
@@ -1016,7 +1016,7 @@
"step": 512.0,
"type": "LLM",
"ui_element": "number",
- "value": 18432,
+ "value": 30208,
"visible": true
},
"llm.max_tokens": {