Avatar for the BerriAI user
BerriAI
litellm
BlogDocsChangelog

feat(responses): add universal server-side context compaction

#25097
Comparing
SwiftWinds:feat/add-compaction
(
2ed929a
) with
main
(
a5322c6
)
CodSpeed Performance Gauge
0%
Untouched
16

Benchmarks

16 total
test_token_counter_long_content
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
+2%
1.7 ms1.7 ms
test_token_counter_raw_text
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
187.1 µs186.2 µs
test_token_counter_with_tools
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
417.7 µs416 µs
test_token_counter_simple_message
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
239 µs238.3 µs
test_get_llm_provider_azure
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
149.2 µs149 µs
test_get_llm_provider_openai
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
143.3 µs143.1 µs
test_get_model_cost_key_exact_match
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
85.1 µs85.1 µs
test_get_model_cost_key_case_insensitive
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
87.4 µs87.5 µs
test_get_llm_provider_with_prefix
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
141.7 µs141.9 µs
test_get_model_info_with_provider
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
85.4 µs85.6 µs
test_cost_per_token_anthropic
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
542.1 µs544.4 µs
test_cost_per_token_openai
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
0%
542 µs544.5 µs
test_get_llm_provider_anthropic
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
-1%
146.6 µs147.5 µs
test_token_counter_multi_turn
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
-1%
569.1 µs574 µs
test_get_model_info_openai
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
-1%
86.6 µs87.5 µs
test_get_model_info_anthropic
tests/benchmarks/test_benchmarks.py
CodSpeed Performance Gauge
-2%
83.7 µs84.9 µs

Commits

Click on a commit to change the comparison range
Base
main
a5322c6
+0.22%
feat(responses): add universal server-side context compaction
3bb4d7e
1 day ago
by SwiftWinds
-0.36%
fix(responses): align compaction format with OpenAI v1/responses spec
6a71bff
23 hours ago
by SwiftWinds
-0.06%
feat: make it not use two user messages in a row
b9c9bdd
22 hours ago
by SwiftWinds
+0.01%
feat: always use our own compaction
330a703
22 hours ago
by SwiftWinds
-0.22%
fix: add greptile fixes
cd56c10
22 hours ago
by SwiftWinds
-0.09%
fix: make it clearer to the llm that a compaction happened
2c4d6cb
21 hours ago
by SwiftWinds
+0.06%
fix: make summarization better and don't use tiktoken
56201bb
21 hours ago
by SwiftWinds
+0.21%
fix: make tests pass after removing created_by
2244997
20 hours ago
by SwiftWinds
0%
feat: add override_native_compaction to make non-native compaction opt-in (if you're using OpenAI)
cef60a1
7 hours ago
by SwiftWinds
-0.24%
fix: do a deep copy instead
0c2aaee
7 hours ago
by SwiftWinds
+0.36%
fix: don't assert created_by is None
2ed929a
7 hours ago
by SwiftWinds
© 2026 CodSpeed Technology
Home Terms Privacy Docs