Skip to content

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

fix: correct cost calculation for cache tokens and simplify memory DevEx

Tejas Chopra787f92520410
diff --git a/headroom/cli.py b/headroom/cli.py
index 78ceedc00..2f2600670 100644
--- a/headroom/cli.py
+++ b/headroom/cli.py
@@ -103,8 +103,8 @@ Usage with OpenAI-compatible clients:
""
if not config.memory_enabled
else '''
-Memory Requirements:
- Set x-headroom-user-id header to identify the user for memory scoping.
+Memory:
+ Memories are scoped per user. Set x-headroom-user-id header for multi-user setups (defaults to 'default').
'''
}
Endpoints:
@@ -488,7 +488,7 @@ Documentation: https://github.com/headroom-sdk/headroom
proxy_parser.add_argument(
"--memory",
action="store_true",
- help="Enable persistent user memory (requires x-headroom-user-id header)",
+ help="Enable persistent user memory (uses x-headroom-user-id header if set, otherwise 'default')",
)
proxy_parser.add_argument(
"--memory-backend",
diff --git a/headroom/proxy/server.py b/headroom/proxy/server.py
index 3a58828a7..ee3bb4502 100644
--- a/headroom/proxy/server.py
+++ b/headroom/proxy/server.py
@@ -1343,22 +1343,10 @@ class HeadroomProxy:
detail=f"Budget exceeded for {self.config.budget_period} period",
)
- # Memory: Validate user ID when memory is enabled
+ # Memory: Get user ID when memory is enabled (fallback to "default" for simple DevEx)
memory_user_id: str | None = None
if self.memory_handler:
- memory_user_id = headers.get("x-headroom-user-id")
- if not memory_user_id:
- return JSONResponse(
- status_code=400,
- content={
- "type": "error",
- "error": {
- "type": "invalid_request_error",
- "message": "x-headroom-user-id header is required when memory is enabled. "
- "Set this header to identify the user for memory scoping.",
- },
- },
- )
+ memory_user_id = headers.get("x-headroom-user-id", "default")
# Check cache (non-streaming only)
cache_hit = False
@@ -3070,7 +3058,9 @@ class HeadroomProxy:
# Use actual tokens from API if available, fallback to estimates
# Note: use 'is not None' instead of 'or' to handle 0 correctly
api_input_tokens = stream_state["input_tokens"]
- total_input_tokens = api_input_tokens if api_input_tokens is not None else optimized_tokens
+ total_input_tokens = (
+ api_input_tokens if api_input_tokens is not None else optimized_tokens
+ )
cache_read_tokens = stream_state["cache_read_input_tokens"]
cache_write_tokens = stream_state["cache_creation_input_tokens"]
diff --git a/tests/test_proxy_memory_integration.py b/tests/test_proxy_memory_integration.py
index 5e88525ee..624060061 100644
--- a/tests/test_proxy_memory_integration.py
+++ b/tests/test_proxy_memory_integration.py
@@ -87,8 +87,8 @@ def anthropic_api_key():
class TestMemoryHeaderValidation:
"""Test user ID header validation."""
- def test_missing_user_id_returns_400(self, memory_client, anthropic_api_key):
- """Request without x-headroom-user-id should return 400 when memory enabled."""
+ def test_missing_user_id_uses_default(self, memory_client, anthropic_api_key):
+ """Request without x-headroom-user-id should use 'default' user for simple DevEx."""
if not anthropic_api_key:
pytest.skip("ANTHROPIC_API_KEY not set")
@@ -97,7 +97,7 @@ class TestMemoryHeaderValidation:
headers={
"x-api-key": anthropic_api_key,
"anthropic-version": "2023-06-01",
- # Note: NOT setting x-headroom-user-id
+ # Note: NOT setting x-headroom-user-id - should default to "default"
},
json={
"model": "claude-sonnet-4-20250514",
@@ -105,10 +105,8 @@ class TestMemoryHeaderValidation:
"messages": [{"role": "user", "content": "Hello"}],
},
)
- assert response.status_code == 400
- error = response.json()
- assert "error" in error
- assert "x-headroom-user-id" in error["error"]["message"]
+ # Should succeed, not return 400
+ assert response.status_code == 200
def test_with_user_id_succeeds(self, memory_client, anthropic_api_key):
"""Request with x-headroom-user-id should succeed."""