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.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
fix(reporting): price the savings tile instead of fabricating $0.00 (#3821)
Aashish Mahatof519fa8b97aa
diff --git a/headroom/reporting/generator.py b/headroom/reporting/generator.pyindex 3484350b0..2a99eec05 100644--- a/headroom/reporting/generator.py+++ b/headroom/reporting/generator.py@@ -7,7 +7,7 @@ from pathlib import Pathfrom typing import TYPE_CHECKING, Anyfrom ..storage import create_storage-from ..utils import estimate_cost, format_cost+from ..utils import format_costif TYPE_CHECKING:pass@@ -337,9 +337,16 @@ def generate_report(tpm_multiplier = 1.0# Estimate cost savings (using gpt-4o pricing)- cost_before = estimate_cost(stats["total_tokens_before"], 0, "gpt-4o") or 0.0- cost_after = estimate_cost(stats["total_tokens_after"], 0, "gpt-4o") or 0.0- estimated_savings = format_cost(cost_before - cost_after)+ # Imported here: the LiteLLM-backed pricer costs ~1.2s of import time.+ from ..models.registry import ModelRegistry++ cost_before = ModelRegistry.estimate_cost("gpt-4o", stats["total_tokens_before"], 0)+ cost_after = ModelRegistry.estimate_cost("gpt-4o", stats["total_tokens_after"], 0)+ # "Cannot price this" is not "$0 saved", so an unpriced report says so.+ if cost_before is None or cost_after is None:+ estimated_savings = "unavailable"+ else:+ estimated_savings = format_cost(cost_before - cost_after)stats["tpm_multiplier"] = tpm_multiplierstats["estimated_savings"] = estimated_savingsdiff --git a/tests/test_reporting.py b/tests/test_reporting.pyindex 9d9f87d36..afac1a5c7 100644--- a/tests/test_reporting.py+++ b/tests/test_reporting.py@@ -9,7 +9,9 @@ from types import SimpleNamespaceimport pytestimport headroom.reporting as reporting+from headroom.models.registry import ModelRegistryfrom headroom.reporting import generator+from headroom.utils import format_cost@dataclass@@ -255,12 +257,6 @@ def test_generate_report_writes_output_and_closes_storage(generator, "_generate_recommendations", lambda *args: [{"title": "Keep going"}])monkeypatch.setattr(generator, "_get_jinja2_template", lambda template_str: FakeTemplate())- monkeypatch.setattr(- generator,- "estimate_cost",- lambda tokens, output_tokens, model: {100: 2.0, 0: None}[tokens],- )- monkeypatch.setattr(generator, "format_cost", lambda cost: f"${cost:.2f}")output_path = tmp_path / "report.html"result = generator.generate_report(@@ -274,10 +270,68 @@ def test_generate_report_writes_output_and_closes_storage(assert output_path.read_text() == "<html>report</html>"assert render_calls[0]["period"] == expected_periodassert render_calls[0]["stats"]["tpm_multiplier"] == 100.0- assert render_calls[0]["stats"]["estimated_savings"] == "$2.00"assert storage.closed is True+def _capture_stats(monkeypatch, tmp_path, stats: dict) -> dict:+ """Run generate_report() over *stats* and hand back the rendered stats dict.++ Only the storage and template seams are patched: pricing stays the product's+ own, so a report that cannot price itself shows up here as a wrong number.+ """+ render_calls: list[dict] = []++ class FakeTemplate:+ def render(self, **kwargs) -> str:+ render_calls.append(kwargs)+ return "<html>report</html>"++ monkeypatch.setattr(generator, "create_storage", lambda store_url: FakeStorage(stats, []))+ monkeypatch.setattr(generator, "_build_waste_histogram", lambda *args: [])+ monkeypatch.setattr(generator, "_get_top_waste_requests", lambda *args, **kwargs: [])+ monkeypatch.setattr(generator, "_generate_recommendations", lambda *args: [])+ monkeypatch.setattr(generator, "_get_jinja2_template", lambda template_str: FakeTemplate())++ generator.generate_report("sqlite:///demo.db", output_path=str(tmp_path / "report.html"))+ return render_calls[0]["stats"]+++SAVINGS_STATS = {+ "total_requests": 2,+ "total_tokens_saved": 245_000,+ "avg_tokens_saved": 122_500,+ "total_tokens_before": 350_000,+ "total_tokens_after": 105_000,+ "avg_cache_alignment": 80,+ "audit_count": 0,+ "optimize_count": 2,+}+++def test_generate_report_prices_the_savings_tile(monkeypatch, tmp_path) -> None:+ """245k tokens saved must cost something, not read as a $0.0000 claim."""+ stats = _capture_stats(monkeypatch, tmp_path, dict(SAVINGS_STATS))++ before = ModelRegistry.estimate_cost("gpt-4o", 350_000, 0)+ after = ModelRegistry.estimate_cost("gpt-4o", 105_000, 0)+ assert before is not None and after is not None, "gpt-4o left the bundled rate card"+ assert stats["estimated_savings"] == format_cost(before - after)+ assert stats["estimated_savings"] != "$0.00"+++def test_generate_report_says_unavailable_when_it_cannot_price(monkeypatch, tmp_path) -> None:+ """An unknown price is a missing number, not zero saved."""+ monkeypatch.setattr(+ ModelRegistry,+ "estimate_cost",+ classmethod(lambda cls, model, *args, **kwargs: None),+ )++ stats = _capture_stats(monkeypatch, tmp_path, dict(SAVINGS_STATS))++ assert stats["estimated_savings"] == "unavailable"++def test_generate_report_closes_storage_when_render_fails(monkeypatch, tmp_path) -> None:storage = FakeStorage({@@ -302,8 +356,6 @@ def test_generate_report_closes_storage_when_render_fails(monkeypatch, tmp_path)monkeypatch.setattr(generator, "_get_top_waste_requests", lambda *args, **kwargs: [])monkeypatch.setattr(generator, "_generate_recommendations", lambda *args: [])monkeypatch.setattr(generator, "_get_jinja2_template", lambda template_str: FakeTemplate())- monkeypatch.setattr(generator, "estimate_cost", lambda *args: 0.0)- monkeypatch.setattr(generator, "format_cost", lambda cost: "$0.00")with pytest.raises(RuntimeError, match="boom"):generator.generate_report("sqlite:///demo.db", output_path=str(tmp_path / "report.html"))