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(settings): read provider, embedder and price facts from the server (#3172)

Raghav Chamadiya6e6d151976ce
diff --git a/packages/api-client/src/types/misc.ts b/packages/api-client/src/types/misc.ts
index 7838812a4..6fb315bb1 100644
--- a/packages/api-client/src/types/misc.ts
+++ b/packages/api-client/src/types/misc.ts
@@ -35,12 +35,28 @@ export interface ProviderInfo {
setup_hint?: string;
}
+export interface EmbedderInfo {
+ id: string;
+ /** Env vars a user sets to make the embedder usable. */
+ env_vars?: string[];
+ /** `false` for the keyless embedder, whose vectors carry no signal. */
+ semantic?: boolean;
+}
+
export interface ProvidersResponse {
active: {
provider: string | null;
model: string | null;
+ /** USD per 1K tokens for `model`; null when no model resolves. */
+ input_cost_per_1k?: number | null;
+ output_cost_per_1k?: number | null;
+ /** The embedder this server runs with. */
+ embedder?: string | null;
};
providers: ProviderInfo[];
+ /** Registerable providers the catalog leaves out. */
+ flag_only_providers?: string[];
+ embedders?: EmbedderInfo[];
}
/** Result of a live provider smoke test. `ok: false` carries the reason in
diff --git a/packages/cli/src/repowise/cli/cost_estimator/__init__.py b/packages/cli/src/repowise/cli/cost_estimator/__init__.py
index e5c49d993..9f25de117 100644
--- a/packages/cli/src/repowise/cli/cost_estimator/__init__.py
+++ b/packages/cli/src/repowise/cli/cost_estimator/__init__.py
@@ -10,16 +10,16 @@ from repowise.core.cost_estimator import (
CostEstimate,
CostRange,
PageTypePlan,
- _lookup_cost,
build_generation_plan,
estimate_cost,
+ lookup_cost,
)
__all__ = [
"CostEstimate",
"CostRange",
"PageTypePlan",
- "_lookup_cost",
"build_generation_plan",
"estimate_cost",
+ "lookup_cost",
]
diff --git a/packages/core/src/repowise/core/cost_estimator/__init__.py b/packages/core/src/repowise/core/cost_estimator/__init__.py
index 52ef39d11..433d54f8e 100644
--- a/packages/core/src/repowise/core/cost_estimator/__init__.py
+++ b/packages/core/src/repowise/core/cost_estimator/__init__.py
@@ -9,13 +9,14 @@ Public API kept stable for existing callers:
- :func:`estimate_cost` — given plans + model, returns a
:class:`CostEstimate` with a median + low/high range. Optionally
calibrated against ``.repowise/db.sqlite`` telemetry.
-- :func:`_lookup_cost` — retained for the provider unit tests.
+- :func:`lookup_cost` — per-1K rates for a model, ``None`` when unpriced;
+ the server prices the active model with it.
"""
from .approx import approximate_generation_plan
from .estimator import STRUCTURAL_PAGE_TYPES, estimate_cost
from .plans import build_generation_plan
-from .pricing import _lookup_cost
+from .pricing import lookup_cost
from .types import CostEstimate, CostRange, PageTypePlan
__all__ = [
@@ -23,8 +24,8 @@ __all__ = [
"CostEstimate",
"CostRange",
"PageTypePlan",
- "_lookup_cost",
"approximate_generation_plan",
"build_generation_plan",
"estimate_cost",
+ "lookup_cost",
]
diff --git a/packages/core/src/repowise/core/cost_estimator/estimator.py b/packages/core/src/repowise/core/cost_estimator/estimator.py
index b09ee79b2..488c29bb6 100644
--- a/packages/core/src/repowise/core/cost_estimator/estimator.py
+++ b/packages/core/src/repowise/core/cost_estimator/estimator.py
@@ -6,7 +6,7 @@ from pathlib import Path
from .calibration import load_telemetry_averages
from .heuristics import HEURISTIC_VARIANCE, heuristic_tokens
-from .pricing import _lookup_cost
+from .pricing import lookup_cost
from .types import CostEstimate, CostRange, PageTypePlan
# Page types rendered from structure. They make no provider call, so they cost
@@ -64,7 +64,8 @@ def estimate_cost(
total_input += inp * plan.count
total_output += out * plan.count
- input_rate, output_rate = _lookup_cost(model_name)
+ # An unpriced model is estimated as free; the CLI gate has always done so.
+ input_rate, output_rate = lookup_cost(model_name) or (0.0, 0.0)
median_cost = (total_input / 1000) * input_rate + (total_output / 1000) * output_rate
# Tighter variance when telemetry calibrated us; wider for cold-start.
diff --git a/packages/core/src/repowise/core/cost_estimator/pricing.py b/packages/core/src/repowise/core/cost_estimator/pricing.py
index 09e3eb7ef..a405a11a3 100644
--- a/packages/core/src/repowise/core/cost_estimator/pricing.py
+++ b/packages/core/src/repowise/core/cost_estimator/pricing.py
@@ -10,9 +10,8 @@ over-estimate on a re-run: safe direction, wrong number.
The token count is not the missing piece: ``cached_tokens`` already flows
provider -> ``GeneratedPage`` -> the ``pages`` table, and the run report
-prints it. What is missing is a third rate (here, in
-``generation/cost_tracker.py``, and in the TS mirror at
-``packages/ui/src/dashboard/quick-actions.tsx``), plus splitting the
+prints it. What is missing is a third rate (here and in
+``generation/cost_tracker.py``), plus splitting the
cached count out of ``input_tokens`` in the cost arithmetic. Worth doing
when we report cost per re-index, where the discount is the whole story;
not worth it to move a one-off init estimate by ~10%.
@@ -54,7 +53,7 @@ _COST_TABLE_EXACT: dict[str, tuple[float, float]] = {
# Prefix fallbacks for unknown variants. No `gpt-5.6` catch-all: the 5.6
# variants are not one price tier, and an unpriced model reads as free here
-# (``_lookup_cost`` falls through to (0.0, 0.0)), so a guess would be worse
+# (the estimator prices a ``None`` from ``lookup_cost`` at zero), so a guess would be worse
# than the miss it hides.
_COST_TABLE_PREFIX: dict[str, tuple[float, float]] = {
"gpt-5.6-luna": (0.0002, 0.0012),
@@ -87,8 +86,12 @@ _COST_TABLE_PREFIX: dict[str, tuple[float, float]] = {
}
-def _lookup_cost(model_name: str) -> tuple[float, float]:
- """Return ``(input_rate, output_rate)`` per 1K tokens for *model_name*."""
+def lookup_cost(model_name: str) -> tuple[float, float] | None:
+ """Return ``(input_rate, output_rate)`` per 1K tokens for *model_name*.
+
+ ``None`` when no entry matches: the model is unpriced, which is not the
+ same as free (the explicit zero rows above).
+ """
lower = model_name.lower()
# OpenRouter/LiteLLM slugs carry a routing prefix (`google/gemini-3.5-flash-lite`)
# that hides the model from every entry below, which priced them at zero.
@@ -98,7 +101,7 @@ def _lookup_cost(model_name: str) -> tuple[float, float]:
if lower in _COST_TABLE_EXACT:
return _COST_TABLE_EXACT[lower]
best_prefix = ""
- best_rates = (0.0, 0.0)
+ best_rates: tuple[float, float] | None = None
for prefix, rates in _COST_TABLE_PREFIX.items():
if lower.startswith(prefix) and len(prefix) > len(best_prefix):
best_prefix = prefix
diff --git a/packages/core/src/repowise/core/providers/embedding/registry.py b/packages/core/src/repowise/core/providers/embedding/registry.py
index 16191404b..91be91041 100644
--- a/packages/core/src/repowise/core/providers/embedding/registry.py
+++ b/packages/core/src/repowise/core/providers/embedding/registry.py
@@ -33,6 +33,14 @@ _BUILTIN_EMBEDDERS: dict[str, tuple[str, str]] = {
"mock": ("repowise.core.providers.embedding.base", "KeylessEmbedder"),
}
+#: What ``REPOWISE_EMBEDDER`` means when unset.
+DEFAULT_EMBEDDER = "mock"
+
+#: Built-ins whose vectors carry no signal (see ``KeylessEmbedder``).
+KEYLESS_EMBEDDERS = frozenset(
+ name for name, (_module, cls) in _BUILTIN_EMBEDDERS.items() if cls == "KeylessEmbedder"
+)
+
_custom_embedders: dict[str, Callable[..., Embedder]] = {}
diff --git a/packages/server/src/repowise/server/app.py b/packages/server/src/repowise/server/app.py
index d219736ad..db7b8a59b 100644
--- a/packages/server/src/repowise/server/app.py
+++ b/packages/server/src/repowise/server/app.py
@@ -35,6 +35,7 @@ from repowise.core.persistence.search import FullTextSearch
from repowise.core.providers.embedding import is_semantic_embedder
from repowise.core.providers.embedding.base import KeylessEmbedder
from repowise.core.providers.embedding.caching import CachingEmbedder
+from repowise.core.providers.embedding.registry import DEFAULT_EMBEDDER
from repowise.server import __version__
from repowise.server.routers import (
actions,
@@ -121,7 +122,7 @@ def _build_embedder():
openrouter — OpenRouterEmbedder via OPENROUTER_API_KEY env var
edenai — EdenAIEmbedder via EDENAI_API_KEY env var
"""
- name = os.environ.get("REPOWISE_EMBEDDER", "mock").lower()
+ name = os.environ.get("REPOWISE_EMBEDDER", DEFAULT_EMBEDDER).lower()
if name == "ollama":
from repowise.core.providers.embedding.ollama import OllamaEmbedder
diff --git a/packages/server/src/repowise/server/provider_config.py b/packages/server/src/repowise/server/provider_config.py
index f26ab8934..462e3b406 100644
--- a/packages/server/src/repowise/server/provider_config.py
+++ b/packages/server/src/repowise/server/provider_config.py
@@ -29,6 +29,11 @@ import os
from pathlib import Path
from typing import Any
+from repowise.core.providers.embedding.registry import (
+ DEFAULT_EMBEDDER,
+ KEYLESS_EMBEDDERS,
+ list_embedders,
+)
from repowise.core.providers.llm.specs import PROVIDER_SPECS
logger = logging.getLogger(__name__)
@@ -57,6 +62,25 @@ PROVIDER_CATALOG: list[dict[str, Any]] = [
_CATALOG_BY_ID = {p["id"]: p for p in PROVIDER_CATALOG}
+FLAG_ONLY_PROVIDERS = [spec.name for spec in PROVIDER_SPECS.values() if spec.picker_rank is None]
+
+
+def _embedder_catalog() -> list[dict[str, Any]]:
+ """Every registered embedder, read at call time so custom ones appear.
+
+ A built-in embedder reads the same credentials as the LLM provider of the
+ same name, so its env vars come from that spec; a custom one has none.
+ The default comes first so the picker opens on it; the rest stay sorted.
+ """
+ return [
+ {
+ "id": name,
+ "env_vars": list(spec.required_envs) if (spec := PROVIDER_SPECS.get(name)) else [],
+ "semantic": name not in KEYLESS_EMBEDDERS,
+ }
+ for name in sorted(list_embedders(), key=lambda n: n != DEFAULT_EMBEDDER)
+ ]
+
# ---------------------------------------------------------------------------
# Config file I/O
@@ -303,13 +327,30 @@ def list_provider_status(
}
)
+ # Deferred: the estimator package pulls in the generation pipeline.
+ from repowise.core.cost_estimator import lookup_cost
+
+ model = active_model or (
+ _CATALOG_BY_ID.get(active_id, {}).get("default_model") if active_id else None
+ )
+ rates = lookup_cost(model) if model else None
+ # An unpriced model on a local provider costs nothing; anywhere else it
+ # is unknown, and null rates make the dashboard show no estimate.
+ spec = PROVIDER_SPECS.get(active_id or "")
+ if rates is None and model and spec is not None and spec.local:
+ rates = (0.0, 0.0)
+ input_rate, output_rate = rates or (None, None)
return {
"active": {
"provider": active_id,
- "model": active_model
- or (_CATALOG_BY_ID.get(active_id, {}).get("default_model") if active_id else None),
+ "model": model,
+ "input_cost_per_1k": input_rate,
+ "output_cost_per_1k": output_rate,
+ "embedder": os.environ.get("REPOWISE_EMBEDDER", DEFAULT_EMBEDDER).lower(),
},
"providers": providers,
+ "flag_only_providers": FLAG_ONLY_PROVIDERS,
+ "embedders": _embedder_catalog(),
}
diff --git a/packages/server/src/repowise/server/schemas/__init__.py b/packages/server/src/repowise/server/schemas/__init__.py
index 5df76167c..7255e92d7 100644
--- a/packages/server/src/repowise/server/schemas/__init__.py
+++ b/packages/server/src/repowise/server/schemas/__init__.py
@@ -262,6 +262,7 @@ if TYPE_CHECKING:
ActiveProviderSelection,
CostGroupResponse,
CostSummaryResponse,
+ EmbedderEntry,
OkResponse,
ProviderEntry,
ProviderStatusResponse,
@@ -544,6 +545,7 @@ _EXPORTS: dict[str, tuple[str, ...]] = {
"ActiveProviderSelection",
"CostGroupResponse",
"CostSummaryResponse",
+ "EmbedderEntry",
"OkResponse",
"ProviderEntry",
"ProviderStatusResponse",
@@ -716,6 +718,7 @@ __all__ = [
"DocDriftResponse",
"DocDriftSummaryResponse",
"EgoGraphResponse",
+ "EmbedderEntry",
"EpisodeCountsResponse",
"EpisodeDetail",
"EpisodeListResponse",
diff --git a/packages/server/src/repowise/server/schemas/ui_helpers.py b/packages/server/src/repowise/server/schemas/ui_helpers.py
index e753b204e..0ef2eb2bc 100644
--- a/packages/server/src/repowise/server/schemas/ui_helpers.py
+++ b/packages/server/src/repowise/server/schemas/ui_helpers.py
@@ -50,16 +50,34 @@ class ProviderEntry(BaseModel):
setup_hint: str = ""
+class EmbedderEntry(BaseModel):
+ """One registered embedder, as the settings picker renders it."""
+
+ id: str
+ #: Env vars a user sets to make the embedder usable.
+ env_vars: list[str] = []
+ #: ``False`` for the keyless embedder, whose vectors carry no signal.
+ semantic: bool = True
+
+
class ActiveProviderSelection(BaseModel):
"""The provider/model this scope resolves to; ``None`` when unset."""
provider: str | None = None
model: str | None = None
+ #: USD per 1K tokens for ``model``; ``None`` when no model resolves.
+ input_cost_per_1k: float | None = None
+ output_cost_per_1k: float | None = None
+ #: The embedder this server runs with.
+ embedder: str | None = None
class ProviderStatusResponse(BaseModel):
active: ActiveProviderSelection
providers: list[ProviderEntry] = []
+ #: Registerable providers the catalog leaves out (picker_rank ``None``).
+ flag_only_providers: list[str] = []
+ embedders: list[EmbedderEntry] = []
class ProviderValidationResponse(BaseModel):
diff --git a/packages/types/src/generated/http.ts b/packages/types/src/generated/http.ts
index 62f0e4acc..4f6e5bd76 100644
--- a/packages/types/src/generated/http.ts
+++ b/packages/types/src/generated/http.ts
@@ -91,6 +91,9 @@ export interface ActionsResponse {
export interface ActiveProviderSelection {
provider?: string | null;
model?: string | null;
+ input_cost_per_1k?: number | null;
+ output_cost_per_1k?: number | null;
+ embedder?: string | null;
}
export interface AgentPromptResponse {
@@ -1266,6 +1269,13 @@ export interface EgoGraphResponse {
outbound_count: number;
}
+/** One registered embedder, as the settings picker renders it. */
+export interface EmbedderEntry {
+ id: string;
+ env_vars?: string[];
+ semantic?: boolean;
+}
+
/**
* Counts by tier and by kind, from a grouped read of the same filters.
*
@@ -2632,6 +2642,8 @@ export interface ProviderEntry {
export interface ProviderStatusResponse {
active: ActiveProviderSelection;
providers?: ProviderEntry[];
+ flag_only_providers?: string[];
+ embedders?: EmbedderEntry[];
}
/**
diff --git a/packages/ui/__tests__/dashboard/quick-actions.test.tsx b/packages/ui/__tests__/dashboard/quick-actions.test.tsx
index 32e93b331..111e7d379 100644
--- a/packages/ui/__tests__/dashboard/quick-actions.test.tsx
+++ b/packages/ui/__tests__/dashboard/quick-actions.test.tsx
@@ -30,6 +30,26 @@ describe("QuickActions", () => {
await waitFor(() => expect(onAction).toHaveBeenCalledWith("sync"));
});
+ it("prices the estimate with the rates it is given, and shows none without them", async () => {
+ const { unmount } = render(
+ <QuickActions
+ onAction={vi.fn()}
+ pageCount={100}
+ modelName="some-model"
+ costPer1k={{ input: 0.001, output: 0.005 }}
+ />,
+ );
+ fireEvent.click(screen.getByRole("button", { name: /Full Re-index/i }));
+ // 100 pages * (3.5K in * $0.001 + 2.2K out * $0.005) = $1.45.
+ expect(await screen.findByText(/1\.45/)).toBeInTheDocument();
+ unmount();
+
+ render(<QuickActions onAction={vi.fn()} pageCount={100} modelName="some-model" />);
+ fireEvent.click(screen.getByRole("button", { name: /Full Re-index/i }));
+ expect(await screen.findByText("Full Re-index", { selector: "h2" })).toBeInTheDocument();
+ expect(screen.queryByText(/Estimated Cost/i)).not.toBeInTheDocument();
+ });
+
it("renders activeJobSlot in place of the buttons when provided", () => {
render(
<QuickActions
diff --git a/packages/ui/src/dashboard/quick-actions.tsx b/packages/ui/src/dashboard/quick-actions.tsx
index 495b0a843..5d20925a5 100644
--- a/packages/ui/src/dashboard/quick-actions.tsx
+++ b/packages/ui/src/dashboard/quick-actions.tsx
@@ -18,56 +18,16 @@ import { formatNumber, formatCost, formatRelativeTime } from "../lib/format";
const AVG_INPUT_TOKENS_PER_PAGE = 3500;
const AVG_OUTPUT_TOKENS_PER_PAGE = 2200;
-const COST_TABLE_EXACT: Record<string, [number, number]> = {
- "gpt-5.4": [0.0025, 0.015],
- "gpt-5.4-mini": [0.00075, 0.0045],
- "gpt-5.4-nano": [0.0002, 0.00125],
- "gpt-5.6-luna": [0.0002, 0.0012],
- "gemini-3.1-pro-preview": [0.002, 0.012],
- "gemini-3-flash-preview": [0.0005, 0.003],
- "gemini-3.1-flash-lite-preview": [0.00025, 0.0015],
- "claude-opus-4-6": [0.005, 0.025],
- "claude-sonnet-4-6": [0.003, 0.015],
- "claude-haiku-4-5": [0.001, 0.005],
- "claude-haiku-5-5": [0.0001, 0.0005],
- "deepseek-v4-flash": [0.00014, 0.00028],
- "deepseek-v4-pro": [0.00174, 0.00348],
-};
-
-const COST_TABLE_PREFIX: [string, [number, number]][] = [
- ["gpt-5.6-luna", [0.0002, 0.0012]],
- ["gpt-5.4-nano", [0.0002, 0.00125]],
- ["gpt-5.4-mini", [0.00075, 0.0045]],
- ["gpt-5.4", [0.0025, 0.015]],
- ["claude-opus", [0.005, 0.025]],
- ["claude-sonnet", [0.003, 0.015]],
- ["claude-haiku", [0.001, 0.005]],
- ["claude", [0.003, 0.015]],
- ["deepseek", [0.00014, 0.00028]],
- ["gemini", [0.00025, 0.0015]],
- ["llama", [0, 0]],
- ["mock", [0, 0]],
-];
-
-function lookupCost(modelName: string): [number, number] {
- const lower = modelName.toLowerCase();
- if (lower in COST_TABLE_EXACT) return COST_TABLE_EXACT[lower]!;
- let bestLen = 0;
- let bestRates: [number, number] = [0, 0];
- for (const [prefix, rates] of COST_TABLE_PREFIX) {
- if (lower.startsWith(prefix) && prefix.length > bestLen) {
- bestLen = prefix.length;
- bestRates = rates;
- }
- }
- return bestRates;
+/** USD per 1K tokens, as the server prices the active model. */
+export interface CostPer1k {
+ input: number;
+ output: number;
}
-function estimateCost(pageCount: number, modelName: string) {
+function estimateCost(pageCount: number, rates: CostPer1k) {
const inputTokens = pageCount * AVG_INPUT_TOKENS_PER_PAGE;
const outputTokens = pageCount * AVG_OUTPUT_TOKENS_PER_PAGE;
- const [inputRate, outputRate] = lookupCost(modelName);
- const cost = (inputTokens / 1000) * inputRate + (outputTokens / 1000) * outputRate;
+ const cost = (inputTokens / 1000) * rates.input + (outputTokens / 1000) * rates.output;
return { inputTokens, outputTokens, cost };
}
@@ -133,6 +93,8 @@ export interface QuickActionsProps {
/** Inputs for the cost-estimate panel inside the confirm dialog. */
pageCount?: number;
modelName?: string;
+ /** The model's rates from the server; without them no estimate is shown. */
+ costPer1k?: CostPer1k | null;
/** When provided, replaces the buttons with the given node — typically a
* rendered job-progress widget. Wrapper decides what to pass. */
activeJobSlot?: ReactNode;
@@ -156,6 +118,7 @@ export function QuickActions({
lastResyncAt,
pageCount = 0,
modelName = "",
+ costPer1k,
activeJobSlot,
variant = "row",
}: QuickActionsProps) {
@@ -165,11 +128,13 @@ export function QuickActions({
const loading = loadingKey !== undefined ? loadingKey : internalLoading;
const estimate = useMemo(() => {
- if (!pendingAction || pendingAction.key === "dead-code" || pageCount <= 0) return null;
+ if (!pendingAction || pendingAction.key === "dead-code" || pageCount <= 0 || !costPer1k) {
+ return null;
+ }
const pages =
pendingAction.key === "sync" ? Math.max(1, Math.ceil(pageCount * 0.1)) : pageCount;
- return estimateCost(pages, modelName);
- }, [pendingAction, pageCount, modelName]);
+ return estimateCost(pages, costPer1k);
+ }, [pendingAction, pageCount, costPer1k]);
async function execute(action: QuickActionDef) {
if (loadingKey === undefined) setInternalLoading(action.key);
diff --git a/packages/web/messages/en.json b/packages/web/messages/en.json
index 2042781ca..1290b4caa 100644
--- a/packages/web/messages/en.json
+++ b/packages/web/messages/en.json
@@ -511,7 +511,7 @@
"embedderLabel": "Embedder",
"embedderHint": "What semantic search is built from. The mock embedder disables it.",
"embedderAria": "Embedder",
- "embedderOffNote": "Semantic search is off. Set <code>REPOWISE_EMBEDDER=gemini</code> or <code>REPOWISE_EMBEDDER=openai</code> on the server for real retrieval.",
+ "embedderOffNote": "Semantic search is off. Pick another embedder above and set <code>REPOWISE_EMBEDDER</code> on the server for real retrieval.",
"embedderOnNote": "Set <code>{env}</code> on the server.",
"catalogUnavailable": "Could not load the provider list from the server, so only the saved provider is shown."
}
diff --git a/packages/web/messages/zh-CN.json b/packages/web/messages/zh-CN.json
index 768aee1e0..895bdbe5f 100644
--- a/packages/web/messages/zh-CN.json
+++ b/packages/web/messages/zh-CN.json
@@ -511,7 +511,7 @@
"embedderLabel": "嵌入模型",
"embedderHint": "语义搜索基于它构建。mock 嵌入模型会禁用语义搜索。",
"embedderAria": "嵌入模型",
- "embedderOffNote": "语义搜索已关闭。请在服务端设置 <code>REPOWISE_EMBEDDER=gemini</code> 或 <code>REPOWISE_EMBEDDER=openai</code> 以启用真实检索。",
+ "embedderOffNote": "语义搜索已关闭。请在上方选择其他嵌入模型,并在服务端设置 <code>REPOWISE_EMBEDDER</code> 以启用真实检索。",
"embedderOnNote": "请在服务端设置 <code>{env}</code>。",
"catalogUnavailable": "无法从服务器加载提供商列表,因此只显示已保存的提供商。"
}
diff --git a/packages/web/src/app/repos/[id]/overview/page.tsx b/packages/web/src/app/repos/[id]/overview/page.tsx
index c32ea3efa..d3ca09f42 100644
--- a/packages/web/src/app/repos/[id]/overview/page.tsx
+++ b/packages/web/src/app/repos/[id]/overview/page.tsx
@@ -88,6 +88,13 @@ export default async function OverviewPage({ params }: Props) {
});
}
+ // Priced by the server; no rates means the dialog shows no estimate.
+ const active = providers?.active;
+ const costPer1k =
+ active?.input_cost_per_1k != null && active.output_cost_per_1k != null
+ ? { input: active.input_cost_per_1k, output: active.output_cost_per_1k }
+ : null;
+
const header = (
<RepoIdentityHeader
name={repo.name}
@@ -103,6 +110,7 @@ export default async function OverviewPage({ params }: Props) {
repoName={repo.name}
pageCount={sync.page_count || stats.file_count}
modelName={providers?.active.model ?? sync.last_sync_model ?? ""}
+ costPer1k={costPer1k}
lastSyncAt={sync.last_sync_at}
lastResyncAt={sync.last_resync_at}
/>
diff --git a/packages/web/src/components/dashboard/quick-actions-wrapper.tsx b/packages/web/src/components/dashboard/quick-actions-wrapper.tsx
index 2ef19894d..1b3951710 100644
--- a/packages/web/src/components/dashboard/quick-actions-wrapper.tsx
+++ b/packages/web/src/components/dashboard/quick-actions-wrapper.tsx
@@ -6,6 +6,7 @@ import { toast } from "sonner";
import {
QuickActions as QuickActionsShell,
DEFAULT_QUICK_ACTIONS,
+ type CostPer1k,
type QuickActionKey,
} from "@repowise-dev/ui/dashboard/quick-actions";
import { syncRepo, fullResyncRepo } from "@/lib/api/repos";
@@ -20,6 +21,7 @@ interface Props {
repoName?: string;
pageCount?: number;
modelName?: string;
+ costPer1k?: CostPer1k | null;
lastSyncAt?: string | null;
lastResyncAt?: string | null;
/** Button arrangement — see `QuickActionsProps.variant`. Overview passes
@@ -37,6 +39,7 @@ export function QuickActionsWrapper({
repoName,
pageCount = 0,
modelName = "",
+ costPer1k,
lastSyncAt,
lastResyncAt,
variant,
@@ -165,6 +168,7 @@ export function QuickActionsWrapper({
lastResyncAt={lastResyncAt}
pageCount={pageCount}
modelName={modelName}
+ costPer1k={costPer1k}
activeJobSlot={activeSlot}
variant={variant}
/>
diff --git a/packages/web/src/components/settings/provider-section.test.tsx b/packages/web/src/components/settings/provider-section.test.tsx
index 26b96f86e..d272f4624 100644
--- a/packages/web/src/components/settings/provider-section.test.tsx
+++ b/packages/web/src/components/settings/provider-section.test.tsx
@@ -7,6 +7,7 @@ import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
const mocks = vi.hoisted(() => ({
getProviders: vi.fn(),
getProvider: vi.fn(() => "gemini"),
+ getEmbedder: vi.fn(() => "mock"),
}));
vi.mock("@/lib/api/providers", () => ({ getProviders: mocks.getProviders }));
@@ -14,7 +15,7 @@ vi.mock("@/lib/config", () => ({
config: {
getProvider: mocks.getProvider,
getModel: () => "",
- getEmbedder: () => "mock",
+ getEmbedder: mocks.getEmbedder,
setProvider: vi.fn(),
setModel: vi.fn(),
setEmbedder: vi.fn(),
@@ -98,6 +99,7 @@ describe("ProviderSection provider catalog", () => {
{ id: "gemini", name: "Google Gemini", default_model: "gemini-3.5-flash-lite" },
{ id: "codex_cli", name: "Codex CLI", default_model: "codex_cli/gpt-5.6-luna" },
],
+ flag_only_providers: ["mock"],
});
render(<ProviderSection />);
@@ -192,3 +194,56 @@ describe("ProviderSection provider catalog", () => {
expect(screen.getByText(/GEMINI_API_KEY/)).toBeTruthy();
});
});
+
+describe("ProviderSection server defaults", () => {
+ beforeEach(() => {
+ vi.clearAllMocks();
+ Element.prototype.hasPointerCapture = vi.fn(() => false);
+ Element.prototype.releasePointerCapture = vi.fn();
+ Element.prototype.scrollIntoView = vi.fn();
+ });
+
+ it("shows the server's provider and embedder when nothing is saved", async () => {
+ mocks.getProvider.mockReturnValue("");
+ mocks.getEmbedder.mockReturnValue("");
+ mocks.getProviders.mockResolvedValue({
+ active: { provider: "openai", model: null, embedder: "gemini" },
+ providers: [{ id: "openai", name: "OpenAI", default_model: "gpt-5.6-luna" }],
+ embedders: [
+ { id: "gemini", env_vars: ["GEMINI_API_KEY"], semantic: true },
+ { id: "mock", env_vars: [], semantic: false },
+ ],
+ });
+
+ render(<ProviderSection />);
+
+ await waitFor(() =>
+ expect(screen.getByRole("combobox", { name: "Provider" }).textContent).toContain("openai"),
+ );
+ expect(screen.getByRole("combobox", { name: "Embedder" }).textContent).toContain("gemini");
+ expect(screen.getByText(/GEMINI_API_KEY/)).toBeTruthy();
+ });
+
+ it("offers the server's embedders and says when one carries no signal", async () => {
+ mocks.getProvider.mockReturnValue("openai");
+ mocks.getEmbedder.mockReturnValue("mock");
+ mocks.getProviders.mockResolvedValue({
+ active: { provider: "openai", model: null, embedder: "mock" },
+ providers: [],
+ embedders: [
+ { id: "voyage", env_vars: [], semantic: true },
+ { id: "mock", env_vars: [], semantic: false },
+ ],
+ });
+
+ render(<ProviderSection />);
+
+ expect(await screen.findByText(/Semantic search is off/)).toBeTruthy();
+ fireEvent.keyDown(screen.getByRole("combobox", { name: "Embedder" }), {
+ key: "Enter",
+ code: "Enter",
+ });
+ await waitFor(() => expect(screen.queryAllByRole("option").length).toBeGreaterThan(0));
+ expect(screen.queryAllByRole("option").map((o) => o.textContent)).toEqual(["voyage", "mock"]);
+ });
+});
diff --git a/packages/web/src/components/settings/provider-section.tsx b/packages/web/src/components/settings/provider-section.tsx
index da316c304..fbe9f311f 100644
--- a/packages/web/src/components/settings/provider-section.tsx
+++ b/packages/web/src/components/settings/provider-section.tsx
@@ -3,7 +3,7 @@
import React, { useEffect, useRef, useState } from "react";
import { config } from "@/lib/config";
import { getProviders } from "@/lib/api/providers";
-import type { ProviderInfo } from "@/lib/api/types";
+import type { EmbedderInfo, ProviderInfo } from "@/lib/api/types";
import { OverviewSection } from "@repowise-dev/ui/overview";
import { Input } from "@repowise-dev/ui/ui/input";
import {
@@ -22,23 +22,6 @@ import {
} from "@repowise-dev/ui/settings";
import { useTranslations } from "next-intl";
-const EMBEDDERS = ["mock", "gemini", "openai", "openrouter", "edenai", "ollama"] as const;
-
-// Real, registerable providers the server catalog deliberately leaves out.
-// `mock` is a keyless test provider (`KEYLESS_PROVIDERS` in the registry) that
-// this page has always offered; it is flag-only, so it is absent from
-// PROVIDER_CATALOG and would otherwise vanish the moment the catalog loads.
-const FLAG_ONLY_PROVIDERS = ["mock"] as const;
-
-const EMBEDDER_ENV_VARS: Record<string, string[]> = {
- gemini: ["GEMINI_API_KEY"],
- openai: ["OPENAI_API_KEY"],
- openrouter: ["OPENROUTER_API_KEY"],
- edenai: ["EDENAI_API_KEY"],
- ollama: ["OLLAMA_BASE_URL"],
- mock: [],
-};
-
/**
* Model and embedder defaults for init/sync triggered from the UI.
*
@@ -51,14 +34,17 @@ const EMBEDDER_ENV_VARS: Record<string, string[]> = {
*/
export function ProviderSection() {
const t = useTranslations("settings");
- const [provider, setProvider] = useState("gemini");
+ // "" until a saved choice or the server's active one is known.
+ const [provider, setProvider] = useState("");
const [model, setModel] = useState("");
- const [embedder, setEmbedder] = useState("mock");
+ const [embedder, setEmbedder] = useState("");
const [serverProvider, setServerProvider] = useState<string | null>(null);
- // The server owns the catalog (it is derived from the provider specs), so
- // this page keeps no copy of its own; until it loads, only the saved
- // provider and the flag-only ones are offered.
+ // The server owns the provider and embedder lists (derived from the core
+ // registries), so this page keeps no copy; until they load, only the saved
+ // choices are offered.
const [catalog, setCatalog] = useState<ProviderInfo[]>([]);
+ const [flagOnly, setFlagOnly] = useState<string[]>([]);
+ const [embedders, setEmbedders] = useState<EmbedderInfo[]>([]);
const [catalogFailed, setCatalogFailed] = useState(false);
const [saveState, setSaveState] = useState<SaveState>("idle");
@@ -70,10 +56,15 @@ export function ProviderSection() {
setEmbedder(config.getEmbedder());
let cancelled = false;
void getProviders()
- .then(({ active, providers: catalog }) => {
+ .then(({ active, providers: catalog, flag_only_providers, embedders: embedderList }) => {
if (cancelled) return;
setServerProvider(active.provider);
setCatalog((catalog ?? []).filter((entry) => entry.id));
+ setFlagOnly(flag_only_providers ?? []);
+ setEmbedders(embedderList ?? []);
+ // Nothing saved: show what the server itself runs with.
+ setProvider((saved) => saved || active.provider || "");
+ setEmbedder((saved) => saved || active.embedder || "");
})
.catch((error: unknown) => {
console.warn("[settings] Could not load the active server provider", error);
@@ -119,11 +110,14 @@ export function ProviderSection() {
// appear in it, and a saved provider the server has since stopped
// advertising would leave a blank trigger with nothing to recover with.
const providerOptions = [
- ...new Set([...catalog.map((entry) => entry.id), ...FLAG_ONLY_PROVIDERS, provider]),
- ];
+ ...new Set([...catalog.map((entry) => entry.id), ...flagOnly, provider]),
+ ].filter(Boolean);
+ const embedderOptions = [
+ ...new Set([...embedders.map((entry) => entry.id), embedder]),
+ ].filter(Boolean);
const providerInfo = catalog.find((entry) => entry.id === provider);
- const embedderVars = EMBEDDER_ENV_VARS[embedder] ?? [];
+ const embedderInfo = embedders.find((entry) => entry.id === embedder);
return (
<OverviewSection
@@ -193,29 +187,29 @@ export function ProviderSection() {
<SelectValue />
</SelectTrigger>
<SelectContent>
- {EMBEDDERS.map((e) => (
+ {embedderOptions.map((e) => (
<SelectItem key={e} value={e}>
{e}
</SelectItem>
))}
</SelectContent>
</Select>
- {embedder === "mock" ? (
+ {embedderInfo?.semantic === false ? (
<EnvVarLine
vars={[]}
note={t.rich("provider.embedderOffNote", {
code: (chunks) => <code className="font-mono text-[var(--color-text-secondary)]">{chunks}</code>,
})}
/>
- ) : (
+ ) : embedderInfo ? (
<EnvVarLine
- vars={embedderVars}
+ vars={embedderInfo.env_vars ?? []}
note={t.rich("provider.embedderOnNote", {
code: (chunks) => <code className="font-mono text-[var(--color-text-secondary)]">{chunks}</code>,
env: `REPOWISE_EMBEDDER=${embedder}`,
})}
/>
- )}
+ ) : null}
</div>
</SettingsRow>
</SettingsRows>
diff --git a/packages/web/src/lib/config.ts b/packages/web/src/lib/config.ts
index 9f6105f25..e8459b44a 100644
--- a/packages/web/src/lib/config.ts
+++ b/packages/web/src/lib/config.ts
@@ -39,13 +39,14 @@ export const config = {
getApiUrl: () => read(KEYS.apiUrl),
setApiUrl: (v: string) => write(KEYS.apiUrl, v),
- getProvider: () => read(KEYS.provider) || "litellm",
+ /** "" when unset; the settings page then shows the server's active one. */
+ getProvider: () => read(KEYS.provider),
setProvider: (v: string) => write(KEYS.provider, v),
getModel: () => read(KEYS.model),
setModel: (v: string) => write(KEYS.model, v),
- getEmbedder: () => read(KEYS.embedder) || "mock",
+ getEmbedder: () => read(KEYS.embedder),
setEmbedder: (v: string) => write(KEYS.embedder, v),
/** Weekend-days preset id; "" means unset, which resolves to Sat/Sun. */
diff --git a/tests/unit/cli/test_pricing_table_agreement.py b/tests/unit/cli/test_pricing_table_agreement.py
index d621516f2..bdb4fd2ac 100644
--- a/tests/unit/cli/test_pricing_table_agreement.py
+++ b/tests/unit/cli/test_pricing_table_agreement.py
@@ -22,7 +22,7 @@ from __future__ import annotations
import pytest
-from repowise.core.cost_estimator.pricing import _COST_TABLE_EXACT, _lookup_cost
+from repowise.core.cost_estimator.pricing import _COST_TABLE_EXACT, lookup_cost
from repowise.core.generation.cost_tracker import _PRICING, get_model_pricing
_SHARED_MODELS = sorted(set(_COST_TABLE_EXACT) & set(_PRICING))
@@ -59,7 +59,7 @@ def test_the_two_tables_actually_overlap() -> None:
@pytest.mark.parametrize("model", [_param(m) for m in _SHARED_MODELS])
def test_estimate_and_live_counter_price_a_model_identically(model: str) -> None:
"""Per-1K x 1000 must equal per-MTok, for both directions of the bill."""
- est_input, est_output = _lookup_cost(model)
+ est_input, est_output = lookup_cost(model)
live = get_model_pricing(model)
assert est_input * 1000 == pytest.approx(live["input"]), (
diff --git a/tests/unit/server/test_provider_config.py b/tests/unit/server/test_provider_config.py
index 33c03e785..a7adbe07b 100644
--- a/tests/unit/server/test_provider_config.py
+++ b/tests/unit/server/test_provider_config.py
@@ -204,7 +204,7 @@ def test_list_provider_status_surfaces_custom_model(clean_env, tmp_path):
status = pc.list_provider_status(repo_id="r1", repo_path=str(repo))
- assert status["active"] == {"provider": "openai", "model": "gemma4"}
+ assert (status["active"]["provider"], status["active"]["model"]) == ("openai", "gemma4")
openai = next(p for p in status["providers"] if p["id"] == "openai")
assert "gemma4" in openai["models"] # custom model is selectable
assert openai["configured"] is True # key seen via repo .env
@@ -315,6 +315,37 @@ def test_list_provider_status_carries_setup_facts_from_the_specs(clean_env, tmp_
assert by_id["codex_cli"]["requires_key"] is False
assert "codex login" in by_id["codex_cli"]["setup_hint"]
assert "mock" not in by_id
+ assert status["flag_only_providers"] == ["mock"]
+
+
+def test_list_provider_status_carries_the_embedder_registry(clean_env, tmp_path, monkeypatch):
+ from repowise.core.providers.embedding import registry
+
+ monkeypatch.setitem(registry._custom_embedders, "voyage", lambda **_kw: None)
+ clean_env["REPOWISE_EMBEDDER"] = "Gemini"
+ status = pc.list_provider_status()
+ by_id = {e["id"]: e for e in status["embedders"]}
+
+ assert set(by_id) == set(registry.list_embedders())
+ assert status["embedders"][0]["id"] == registry.DEFAULT_EMBEDDER
+ assert by_id["gemini"]["env_vars"] == ["GEMINI_API_KEY", "GOOGLE_API_KEY"]
+ assert by_id["mock"] == {"id": "mock", "env_vars": [], "semantic": False}
+ assert by_id["voyage"] == {"id": "voyage", "env_vars": [], "semantic": True}
+ assert status["active"]["embedder"] == "gemini"
+
+
+def test_list_provider_status_prices_the_active_model(clean_env, tmp_path):
+ from repowise.core.cost_estimator import lookup_cost
+
+ def rates(provider, model):
+ pc.set_active_provider(provider, model)
+ active = pc.list_provider_status()["active"]
+ return active["input_cost_per_1k"], active["output_cost_per_1k"]
+
+ assert rates("openai", "gpt-5.4-mini") == lookup_cost("gpt-5.4-mini")
+ # Unpriced is unknown, not free; a local model is free.
+ assert rates("openai", "gpt-4o") == (None, None)
+ assert rates("ollama", "qwen3.5:4b") == (0.0, 0.0)
def test_list_provider_status_never_returns_key_material(clean_env, tmp_path):
diff --git a/tests/unit/test_providers/test_claude_cli_provider.py b/tests/unit/test_providers/test_claude_cli_provider.py
index b009a4fc3..43bf1b38a 100644
--- a/tests/unit/test_providers/test_claude_cli_provider.py
+++ b/tests/unit/test_providers/test_claude_cli_provider.py
@@ -576,12 +576,12 @@ def test_registered_as_keyless_provider():
def test_subscription_usage_is_priced_at_zero():
"""A seat is not API spend; the bare `claude` prefixes would price it wrongly."""
- from repowise.core.cost_estimator.pricing import _lookup_cost
+ from repowise.core.cost_estimator.pricing import lookup_cost
- assert _lookup_cost("claude_cli/claude-haiku-4-5") == (0.0, 0.0)
- assert _lookup_cost("claude_cli/claude-opus-4-6") == (0.0, 0.0)
+ assert lookup_cost("claude_cli/claude-haiku-4-5") == (0.0, 0.0)
+ assert lookup_cost("claude_cli/claude-opus-4-6") == (0.0, 0.0)
# The keyed API path is unaffected.
- assert _lookup_cost("claude-haiku-4-5") != (0.0, 0.0)
+ assert lookup_cost("claude-haiku-4-5") != (0.0, 0.0)
def test_resolves_through_the_registry(claude_on_path):
diff --git a/tests/unit/test_providers/test_cost_estimator.py b/tests/unit/test_providers/test_cost_estimator.py
index cfd309e7b..8dbdaa4fd 100644
--- a/tests/unit/test_providers/test_cost_estimator.py
+++ b/tests/unit/test_providers/test_cost_estimator.py
@@ -4,7 +4,7 @@ from __future__ import annotations
import pytest
-from repowise.cli.cost_estimator import PageTypePlan, _lookup_cost, estimate_cost
+from repowise.cli.cost_estimator import PageTypePlan, estimate_cost, lookup_cost
from repowise.cli.cost_estimator.heuristics import heuristic_tokens
# ---------------------------------------------------------------------------
@@ -19,8 +19,8 @@ from repowise.cli.cost_estimator.heuristics import heuristic_tokens
("gpt-5.4-nano", 0.0002, 0.00125),
("gpt-5.4-mini", 0.00075, 0.0045),
("gpt-5.4", 0.0025, 0.015),
- # The default. An unpriced model reads as free here (_lookup_cost
- # falls through to (0.0, 0.0)), so this row is the guard.
+ # The default. An unpriced model is estimated as free (lookup_cost
+ # returns None), so this row is the guard.
("gpt-5.6-luna", 0.0002, 0.0012),
# Gemini
("gemini-3.1-flash-lite-preview", 0.00025, 0.0015),
@@ -39,7 +39,7 @@ from repowise.cli.cost_estimator.heuristics import heuristic_tokens
],
)
def test_lookup_cost(model, expected_input, expected_output):
- inp, out = _lookup_cost(model)
+ inp, out = lookup_cost(model)
assert inp == pytest.approx(expected_input, rel=1e-6)
assert out == pytest.approx(expected_output, rel=1e-6)
diff --git a/tests/unit/test_providers/test_provider_specs.py b/tests/unit/test_providers/test_provider_specs.py
index fff2371e0..cd80fc3e4 100644
--- a/tests/unit/test_providers/test_provider_specs.py
+++ b/tests/unit/test_providers/test_provider_specs.py
@@ -10,6 +10,7 @@ replaced drifted.
from __future__ import annotations
import ast
+import re
import types
from pathlib import Path
@@ -197,3 +198,49 @@ def test_the_guard_is_looking_at_real_sources():
"""A glob that matched nothing, or a name set that emptied, passes vacuously."""
assert len(_sources()) > 500
assert {"claude_cli", "codex_cli", "opencode", "litellm", "openai"} <= _CHECKED
+
+
+# --- no provider or embedder lists in the web UI ----------------------------
+
+#: A quoted provider/embedder name, or one used as an object key. The UI reads
+#: these from ``/api/providers``; ``mock`` is checked here, unlike above.
+_UI_NAME = re.compile(
+ r"""["'](?P<q>{names})["']|^\s*(?P<k>{names})\s*:""".format(
+ names="|".join(sorted(frozenset(PROVIDER_SPECS) | _EMBEDDERS, key=len, reverse=True))
+ ),
+ re.MULTILINE,
+)
+
+
+def _ui_sources():
+ return sorted(
+ p
+ for pkg in ("ui", "web", "api-client")
+ for p in (_PACKAGES / pkg / "src").rglob("*.ts*")
+ if p.suffix in {".ts", ".tsx"}
+ and ".test." not in p.name
+ and "__tests__" not in p.parts
+ and "generated" not in p.parts
+ )
+
+
+def test_no_ui_source_spells_a_provider_name():
+ offenders = [
+ f"{path.relative_to(_PACKAGES).as_posix()}:"
+ f"{text.count(chr(10), 0, m.start()) + 1} {m.group('q') or m.group('k')!r}"
+ for path in _ui_sources()
+ for text in [path.read_text(encoding="utf-8")]
+ for m in _UI_NAME.finditer(text)
+ ]
+ assert not offenders, f"provider names spelled in UI source; read /api/providers: {offenders}"
+
+
+ "snippet", ['x === "mock"', "['gemini', 'openai']", " ollama: [],", 'useState("litellm")']
+)
+def test_the_ui_guard_sees_each_shape(snippet):
+ assert _UI_NAME.search(snippet)
+
+
+def test_the_ui_guard_is_looking_at_real_sources():
+ assert len(_ui_sources()) > 200