Skip to content

The open-source agent harness - the runtime layer that turns an LLM into a working agent.

Pin upper bounds for pip packages

Chirag Jain402fe596414e
diff --git a/benchmark/requirements-deepagents.txt b/benchmark/requirements-deepagents.txt
index c4b316772..902cafab6 100644
--- a/benchmark/requirements-deepagents.txt
+++ b/benchmark/requirements-deepagents.txt
@@ -1,5 +1,5 @@
# deepagents arm only — install in a separate venv from the core stack.
-deepagents>=0.1
-langgraph>=0.2
-langchain-openai>=0.2
-langchain-mcp-adapters>=0.1
+deepagents>=0.1,<0.2
+langgraph>=0.2,<0.3
+langchain-openai>=0.2,<0.3
+langchain-mcp-adapters>=0.1,<0.2
diff --git a/benchmark/requirements.txt b/benchmark/requirements.txt
index baa3f71f3..028476b6d 100644
--- a/benchmark/requirements.txt
+++ b/benchmark/requirements.txt
@@ -1,7 +1,7 @@
# Core arms (TrueForge + CMA) and the judge.
# - TrueForge arm calls its own /api/v1 HTTP API over the stdlib (no extra dep).
# - CMA arm and the judge use the Anthropic SDK.
-anthropic>=0.40
+anthropic>=0.40,<0.41
# The deepagents arm pulls in LangGraph + LangChain, which can conflict with the
# core stack. If pip resolves a conflict, install the deepagents arm in its own