Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

fix(costs): nano-model pricing + lead the Costs page with agent savings (#422)

Raghav Chamadiyad94dd8a07dc9
diff --git a/.github/assets/savings.png b/.github/assets/savings.png
new file mode 100644
index 000000000..dcb7ccfae
Binary files /dev/null and b/.github/assets/savings.png differ
diff --git a/.github/assets/webui.gif b/.github/assets/webui.gif
index d91b66fd5..fece7ce63 100644
Binary files a/.github/assets/webui.gif and b/.github/assets/webui.gif differ
diff --git a/README.md b/README.md
index 825992072..9d83e6b96 100644
--- a/README.md
+++ b/README.md
@@ -164,6 +164,11 @@ from distilled output as from raw. Opt-in Claude Code hook rewrites noisy
commands automatically (shown for approval); `repowise saved` tracks tokens
and dollars saved. Full guide: **[docs/DISTILL.md →](docs/DISTILL.md)**
+<div align="center">
+<img src=".github/assets/savings.png" alt="repowise Costs dashboard — tokens and dollars saved across distill and MCP tools" width="100%" />
+<p align="center"><sub>The <strong>Costs</strong> dashboard tallies both savings surfaces — <code>repowise distill</code> (command output) and the MCP tools (each curated answer replacing the raw file reads it stood in for) — priced at your coding agent's own model. Example shown from a week of heavy local use.</sub></p>
+</div>
+
### 3 · Code health predicts real defects
Health scores are collected at a historical commit (T0); bug-fixing commits are
diff --git a/packages/core/src/repowise/core/generation/cost_tracker.py b/packages/core/src/repowise/core/generation/cost_tracker.py
index d02c08c3d..4994de761 100644
--- a/packages/core/src/repowise/core/generation/cost_tracker.py
+++ b/packages/core/src/repowise/core/generation/cost_tracker.py
@@ -33,6 +33,11 @@ _PRICING: dict[str, dict[str, float]] = {
"gpt-5": {"input": 1.25, "output": 10.0},
"gpt-5-codex": {"input": 1.25, "output": 10.0},
"gpt-5-mini": {"input": 0.25, "output": 2.0},
+ # Nano tier — $0.05/$0.40 per 1M. Without these keys the model name falls
+ # through to ``_FALLBACK_PRICING`` (Sonnet $3/$15) and overstates a nano
+ # indexing run ~40×.
+ "gpt-5-nano": {"input": 0.05, "output": 0.40},
+ "gpt-5.4-nano": {"input": 0.05, "output": 0.40},
"gpt-4o": {"input": 2.5, "output": 10.0},
"gpt-4o-mini": {"input": 0.15, "output": 0.6},
# Google Gemini
diff --git a/packages/ui/src/costs/distill-savings-card.tsx b/packages/ui/src/costs/distill-savings-card.tsx
index 3875d24bd..912ff6698 100644
--- a/packages/ui/src/costs/distill-savings-card.tsx
+++ b/packages/ui/src/costs/distill-savings-card.tsx
@@ -103,8 +103,10 @@ export function DistillSavingsCard({ data }: DistillSavingsCardProps) {
const distillPct = total > 0 ? Math.round((distillSaved / total) * 100) : 0;
const mcpPct = 100 - distillPct;
- const topFilters = data.per_filter.slice(0, 3);
- const topTools = (data.mcp_per_tool ?? []).slice(0, 3);
+ const topFilters = [...data.per_filter]
+ .sort((a, b) => b.saved_tokens - a.saved_tokens)
+ .slice(0, 5);
+ const topTools = (data.mcp_per_tool ?? []).slice(0, 5);
const priced = agentLabel(data.pricing_agent) || data.pricing_model;
const detected = data.pricing_source && data.pricing_source !== "default";
const missed = (data.missed_tokens_est ?? 0) > 0;
diff --git a/packages/web/src/app/repos/[id]/costs/page.tsx b/packages/web/src/app/repos/[id]/costs/page.tsx
index 5d5dafb53..1ebb823e9 100644
--- a/packages/web/src/app/repos/[id]/costs/page.tsx
+++ b/packages/web/src/app/repos/[id]/costs/page.tsx
@@ -30,7 +30,7 @@ import { formatCost, formatNumber, formatTokens } from "@repowise-dev/ui/lib/for
export default function CostsPage() {
const params = useParams<{ id: string }>();
const id = params.id;
- const [tab, setTab] = useState("daily");
+ const [tab, setTab] = useState("operations");
const { data: summary, isLoading: loadingSummary } = useSWR<CostSummary>(
`costs-summary:${id}`,
@@ -70,7 +70,7 @@ export default function CostsPage() {
Cost Tracking
</h1>
<p className="text-sm text-[var(--color-text-secondary)]">
- LLM token usage and spend across all generation runs.
+ What repowise saved your coding agent — and what generating the docs cost.
</p>
</div>
@@ -78,45 +78,13 @@ export default function CostsPage() {
coding agent, across distill (CLI + hook) and MCP tool responses. */}
<DistillSavingsCard data={distillSavings} />
- {loadingSummary ? (
- <div className="grid grid-cols-2 gap-3 sm:grid-cols-4">
- {Array.from({ length: 4 }).map((_, i) => (
- <Skeleton key={i} className="h-24 w-full rounded-lg" />
- ))}
- </div>
- ) : summary ? (
- <div className="grid grid-cols-2 gap-3 sm:grid-cols-4">
- <StatCard
- label="Total Cost"
- value={formatCost(summary.total_cost_usd)}
- description="all time"
- icon={<DollarSign className="h-4 w-4 text-green-500" />}
- />
- <StatCard
- label="Total Calls"
- value={formatNumber(summary.total_calls)}
- description="LLM API calls"
- />
- <StatCard
- label="Input Tokens"
- value={formatTokens(summary.total_input_tokens)}
- description="prompt tokens"
- />
- <StatCard
- label="Output Tokens"
- value={formatTokens(summary.total_output_tokens)}
- description="completion tokens"
- />
- </div>
- ) : null}
-
<Tabs value={tab} onValueChange={setTab} className="w-full">
<TabsList>
+ <TabsTrigger value="operations">Spend by operation</TabsTrigger>
<TabsTrigger value="daily">Daily</TabsTrigger>
- <TabsTrigger value="cache">Cache</TabsTrigger>
- <TabsTrigger value="hotspots">Hotspots</TabsTrigger>
<TabsTrigger value="providers">Providers</TabsTrigger>
- <TabsTrigger value="operations">Operations</TabsTrigger>
+ <TabsTrigger value="hotspots">Hotspots</TabsTrigger>
+ <TabsTrigger value="cache">Cache</TabsTrigger>
</TabsList>
<TabsContent value="daily" className="mt-4">
@@ -253,6 +221,45 @@ export default function CostsPage() {
</Card>
</TabsContent>
</Tabs>
+
+ {/* Indexing / generation cost — deliberately secondary to the savings
+ hero above. */}
+ <div className="space-y-2">
+ <p className="text-xs font-medium uppercase tracking-wide text-[var(--color-text-tertiary)]">
+ Indexing &amp; generation cost
+ </p>
+ {loadingSummary ? (
+ <div className="grid grid-cols-2 gap-3 sm:grid-cols-4">
+ {Array.from({ length: 4 }).map((_, i) => (
+ <Skeleton key={i} className="h-24 w-full rounded-lg" />
+ ))}
+ </div>
+ ) : summary ? (
+ <div className="grid grid-cols-2 gap-3 sm:grid-cols-4">
+ <StatCard
+ label="Indexing cost"
+ value={formatCost(summary.total_cost_usd)}
+ description="across all generation runs"
+ icon={<DollarSign className="h-4 w-4 text-green-500" />}
+ />
+ <StatCard
+ label="Total Calls"
+ value={formatNumber(summary.total_calls)}
+ description="LLM API calls"
+ />
+ <StatCard
+ label="Input Tokens"
+ value={formatTokens(summary.total_input_tokens)}
+ description="prompt tokens"
+ />
+ <StatCard
+ label="Output Tokens"
+ value={formatTokens(summary.total_output_tokens)}
+ description="completion tokens"
+ />
+ </div>
+ ) : null}
+ </div>
</div>
);
}
diff --git a/tests/unit/generation/test_cost_pricing.py b/tests/unit/generation/test_cost_pricing.py
new file mode 100644
index 000000000..3b2e6298d
--- /dev/null
+++ b/tests/unit/generation/test_cost_pricing.py
@@ -0,0 +1,23 @@
+"""Pricing-table lookups for the cost tracker.
+
+Regression guard for the nano tier: an unlisted model silently falls back to
+the Sonnet-tier default rate, which overstates a cheap indexing run by ~40x.
+"""
+
+from __future__ import annotations
+
+from repowise.core.generation.cost_tracker import (
+ _FALLBACK_PRICING,
+ get_model_pricing,
+)
+
+
+def test_nano_models_priced_at_nano_rate_not_fallback() -> None:
+ for model in ("gpt-5-nano", "gpt-5.4-nano"):
+ pricing = get_model_pricing(model)
+ assert pricing == {"input": 0.05, "output": 0.40}
+ assert pricing != _FALLBACK_PRICING
+
+
+def test_unknown_model_still_falls_back() -> None:
+ assert get_model_pricing("totally-made-up-model-9000") == _FALLBACK_PRICING