diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index 53db06d..42981b0 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -22,4 +22,4 @@ jobs: run: uv sync --extra dev - name: Run tests - run: uv run pytest tests/test_models.py -v + run: uv run pytest -v -m "not integration" diff --git a/docker-compose.yml b/docker-compose.yml index b05c701..60e8645 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -14,7 +14,7 @@ services: API_PORT: ${API_PORT:-8000} ANTHROPIC_API_KEY: ${ANTHROPIC_API_KEY} POLICYENGINE_API_URL: http://localhost:${API_PORT:-8000} - AGENT_USE_MODAL: "false" + AGENT_USE_MODAL: ${AGENT_USE_MODAL:-false} volumes: - ./src:/app/src - ./docs/out:/app/docs/out diff --git a/docs/src/components/policy-chat.tsx b/docs/src/components/policy-chat.tsx index 2a9865b..5a5b101 100644 --- a/docs/src/components/policy-chat.tsx +++ b/docs/src/components/policy-chat.tsx @@ -1,6 +1,6 @@ "use client"; -import { useState, useRef, useEffect } from "react"; +import { useState, useRef, useEffect, useMemo } from "react"; import ReactMarkdown from "react-markdown"; import remarkBreaks from "remark-breaks"; import { useApi } from "./api-context"; @@ -16,6 +16,239 @@ interface LogEntry { message: string; } +interface ParsedStep { + type: "agent" | "tool_use" | "api_call" | "api_response" | "tool_result" | "assistant" | "unknown"; + title: string; + content: string; + method?: string; + url?: string; + statusCode?: number; + toolName?: string; + params?: Record; + isExpanded?: boolean; +} + +function parseLogEntry(message: string): ParsedStep { + // [AGENT] messages - filter out internal debug info + if (message.startsWith("[AGENT]")) { + const content = message.replace("[AGENT] ", ""); + // Skip internal debug messages + if (content.startsWith("Stop reason:") || + content.startsWith("Turn ") || + content.startsWith("Loaded ") || + content.startsWith("Fetching ") || + content.startsWith("Completed")) { + return { type: "unknown", title: "", content: "" }; + } + return { + type: "agent", + title: "Agent", + content: content, + }; + } + + // [TOOL_USE] tool_name: {...} + if (message.startsWith("[TOOL_USE]")) { + const content = message.replace("[TOOL_USE] ", ""); + const colonIndex = content.indexOf(":"); + if (colonIndex > -1) { + const toolName = content.slice(0, colonIndex).trim(); + const paramsStr = content.slice(colonIndex + 1).trim(); + let params: Record = {}; + try { + params = JSON.parse(paramsStr); + } catch { + // Not valid JSON + } + // Clean up tool name for display + const displayName = toolName + .replace(/_/g, " ") + .replace(/parameters get$/, "") + .replace(/parameters post$/, "") + .replace(/household calculate post$/, "Calculate household") + .replace(/list /g, "Search "); + return { + type: "tool_use", + title: displayName, + content: paramsStr, + toolName, + params, + }; + } + } + + // [API] GET/POST url + if (message.startsWith("[API]")) { + const content = message.replace("[API] ", ""); + + // Check if it's a response + if (content.startsWith("Response:")) { + const statusCode = parseInt(content.replace("Response: ", ""), 10); + return { + type: "api_response", + title: "Response", + content: content, + statusCode, + }; + } + + // Check if it's a request with method + const methodMatch = content.match(/^(GET|POST|PUT|PATCH|DELETE)\s+(.+)$/); + if (methodMatch) { + return { + type: "api_call", + title: "API Request", + content: content, + method: methodMatch[1], + url: methodMatch[2], + }; + } + + // Query or Body + if (content.startsWith("Query:") || content.startsWith("Body:")) { + return { + type: "api_call", + title: content.startsWith("Query:") ? "Query params" : "Request body", + content: content.replace(/^(Query|Body):\s*/, ""), + }; + } + } + + // [TOOL_RESULT] ... + if (message.startsWith("[TOOL_RESULT]")) { + const content = message.replace("[TOOL_RESULT] ", ""); + return { + type: "tool_result", + title: "Result", + content: content, + }; + } + + // [ASSISTANT] ... + if (message.startsWith("[ASSISTANT]")) { + const content = message.replace("[ASSISTANT] ", ""); + return { + type: "assistant", + title: "Thinking", + content: content, + }; + } + + return { + type: "unknown", + title: "Log", + content: message, + }; +} + +function ToolCard({ step }: { step: ParsedStep }) { + const [isExpanded, setIsExpanded] = useState(false); + + if (step.type === "agent") { + return null; // Hide agent messages, they're redundant with progress indicator + } + + if (step.type === "tool_use") { + return ( +
+ + {isExpanded && step.params && Object.keys(step.params).length > 0 && ( +
+ {Object.entries(step.params).map(([key, value]) => ( +
+ {key}: + + {typeof value === "string" ? value : JSON.stringify(value)} + +
+ ))} +
+ )} +
+ ); + } + + // Hide API details - too noisy + if (step.type === "api_call" || step.type === "api_response") { + return null; + } + + if (step.type === "tool_result") { + return ( +
+ + {isExpanded && ( +
+ 
{step.content}
+
+ )} +
+ ); + } + + if (step.type === "assistant") { + return ( +
+

{step.content}

+
+ ); + } + + return null; +} + +function ProgressIndicator({ logs }: { logs: LogEntry[] }) { + const stage = useMemo(() => { + const hasSearch = logs.some(l => l.message.includes("parameters")); + const hasPolicy = logs.some(l => l.message.includes("policies")); + const hasAnalysis = logs.some(l => l.message.includes("analysis") || l.message.includes("economic")); + const hasHousehold = logs.some(l => l.message.includes("household")); + const isComplete = logs.some(l => l.message.includes("Completed")); + + if (isComplete) return "Complete"; + if (hasAnalysis) return "Running analysis..."; + if (hasPolicy) return "Creating policy..."; + if (hasHousehold) return "Calculating..."; + if (hasSearch) return "Searching parameters..."; + return "Starting..."; + }, [logs]); + + if (logs.length === 0) return null; + + return ( +
+ {stage !== "Complete" && ( +
+ )} + {stage} +
+ ); +} + export function PolicyChat() { const { baseUrl } = useApi(); const [messages, setMessages] = useState([]); @@ -34,7 +267,6 @@ export function PolicyChat() { scrollToBottom(); }, [messages, logs]); - // Cleanup polling on unmount useEffect(() => { return () => { if (pollIntervalRef.current) { @@ -43,20 +275,21 @@ export function PolicyChat() { }; }, []); + const parsedSteps = useMemo(() => { + return logs + .map(log => parseLogEntry(log.message)) + .filter(step => step.type !== "unknown"); + }, [logs]); + const pollLogs = async (id: string) => { try { const res = await fetch(`${baseUrl}/agent/logs/${id}`); - if (!res.ok) { - console.error("Failed to fetch logs:", res.status); - return; - } + if (!res.ok) return; const data = await res.json(); setLogs(data.logs || []); - // Check if completed or failed if (data.status === "completed" || data.status === "failed") { - // Stop polling if (pollIntervalRef.current) { clearInterval(pollIntervalRef.current); pollIntervalRef.current = null; @@ -65,39 +298,16 @@ export function PolicyChat() { setIsLoading(false); setCallId(null); - // Extract final result from logs or result field let finalContent = ""; if (data.result?.result) { finalContent = data.result.result; } else { - // Try to extract from logs - look for [CLAUDE] lines with result - const claudeLogs = data.logs - .map((l: LogEntry) => l.message) - .filter((m: string) => m.startsWith("[CLAUDE]")) - .map((m: string) => m.replace("[CLAUDE] ", "")); - - // Try to parse the last few lines for result - for (const log of claudeLogs.reverse()) { - try { - const event = JSON.parse(log); - if (event.type === "result" && event.result) { - finalContent = event.result; - break; - } - } catch { - // Not JSON, skip - } - } - - if (!finalContent) { - finalContent = - data.status === "completed" - ? "Analysis completed. Check logs for details." - : "Analysis failed. Check logs for errors."; - } + finalContent = + data.status === "completed" + ? "Analysis completed. Check the steps above for details." + : "Analysis failed. Please try again."; } - // Update assistant message with final content setMessages((prev) => { const newMessages = [...prev]; const lastIndex = newMessages.length - 1; @@ -126,38 +336,30 @@ export function PolicyChat() { setLogs([]); setCallId(null); - // Stop any existing polling if (pollIntervalRef.current) { clearInterval(pollIntervalRef.current); pollIntervalRef.current = null; } - // Add user message setMessages((prev) => [...prev, { role: "user", content: userMessage }]); - - // Add pending assistant message setMessages((prev) => [ ...prev, { role: "assistant", content: "", status: "pending" }, ]); try { - // Start the agent const res = await fetch(`${baseUrl}/agent/run`, { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({ question: userMessage }), }); - if (!res.ok) { - throw new Error(`HTTP ${res.status}`); - } + if (!res.ok) throw new Error(`HTTP ${res.status}`); const data = await res.json(); const newCallId = data.call_id; setCallId(newCallId); - // Update to running status setMessages((prev) => { const newMessages = [...prev]; const lastIndex = newMessages.length - 1; @@ -170,12 +372,10 @@ export function PolicyChat() { return newMessages; }); - // Start polling for logs pollIntervalRef.current = setInterval(() => { pollLogs(newCallId); }, 1000); - // Initial poll pollLogs(newCallId); } catch (err) { setMessages((prev) => { @@ -194,195 +394,186 @@ export function PolicyChat() { } }; - // Parse log message to extract useful info - const parseLogMessage = (message: string): { type: string; content: string } => { - if (message.startsWith("[AGENT]")) { - return { type: "agent", content: message.replace("[AGENT] ", "") }; - } - if (message.startsWith("[CLAUDE]")) { - const claudeContent = message.replace("[CLAUDE] ", ""); - // Try to parse as JSON - try { - const event = JSON.parse(claudeContent); - if (event.type === "assistant" && event.message?.content) { - const textParts = event.message.content - .filter((c: { type: string }) => c.type === "text") - .map((c: { text: string }) => c.text) - .join(""); - if (textParts) { - return { type: "text", content: textParts }; - } - const toolParts = event.message.content - .filter((c: { type: string }) => c.type === "tool_use") - .map((c: { name: string }) => c.name); - if (toolParts.length > 0) { - return { type: "tool", content: `Using: ${toolParts.join(", ")}` }; - } - } - if (event.type === "system" && event.subtype === "init") { - const mcpStatus = event.mcp_servers?.find( - (s: { name: string }) => s.name === "policyengine" - ); - return { - type: "system", - content: mcpStatus?.status === "connected" ? "MCP connected" : "Starting...", - }; - } - if (event.type === "result") { - return { type: "result", content: "Analysis complete" }; - } - return { type: "claude", content: `[${event.type || "event"}]` }; - } catch { - return { type: "claude", content: claudeContent.slice(0, 100) }; - } - } - return { type: "log", content: message.slice(0, 100) }; - }; - const exampleQuestions = [ - "How much would it cost to set the UK basic income tax rate to 19p?", - "What would happen if we doubled child benefit?", - "Calculate tax for a UK household earning 50,000", - "What is the budgetary impact of abolishing the higher rate of income tax?", - "What benefits would a single parent with two children receive in California?", + "What is the UK personal allowance for 2026?", + "Calculate tax for someone earning £50,000 in the UK", + "What would happen if we increased child benefit by 10%?", + "What benefits would a single parent with two children receive?", ]; return ( -
+
{/* Header */} -
-
-
- - Policy analyst - - - Powered by Claude Code + MCP - +
+
+
+
+ + + +
+
+

Policy analyst

+

Ask questions about UK and US tax-benefit policy

+
+
+
+
+ + {isLoading ? "Working..." : "Ready"} + +
-

- Ask natural language questions about UK or US tax and benefit policy -

{/* Messages */} -
- {messages.length === 0 && ( -
-

- Try asking a question like: -

-
+
+ {messages.length === 0 ? ( +
+
+

+ What would you like to know? +

+

+ Ask about tax rates, benefits, or policy impacts +

+
+
{exampleQuestions.map((q, i) => ( ))}
- )} - - {messages.map((message, i) => ( -
-
- {message.role === "assistant" && - (message.status === "pending" || message.status === "running") ? ( -
-
- - {message.status === "pending" ? "Starting..." : "Analysing..."} - -
- ) : message.status === "completed" || message.status === "failed" ? ( -
- - {message.content} - -
- ) : ( -
{message.content}
- )} -
-
- ))} - - {/* Live logs */} - {isLoading && logs.length > 0 && ( -
-
- Live output ({logs.length} entries) -
- {logs.slice(-30).map((log, i) => { - const parsed = parseLogMessage(log.message); - return ( -
- - {">"} - - {parsed.content} -
- ); - })} -
- {">"} - -
+ ) : ( +
+ {messages.map((message, i) => ( +
+ {message.role === "user" ? ( +
+
+

{message.content}

+
+
+ ) : ( +
+ {/* Running state with live steps */} + {(message.status === "pending" || message.status === "running") && ( +
+ + + {message.status === "pending" ? ( +
+
+ Starting analysis... +
+ ) : ( +
+ {parsedSteps.slice(-10).map((step, j) => ( + + ))} +
+ )} +
+ )} + + {/* Completed/failed state */} + {(message.status === "completed" || message.status === "failed") && ( +
+ {/* Collapsible steps summary */} + {parsedSteps.length > 0 && ( +
+ + + + + {parsedSteps.filter(s => s.type === "tool_use").length} tool calls executed + +
+ {parsedSteps.map((step, j) => ( + + ))} +
+
+ )} + + {/* Final response */} +
+
+ + {message.content} + +
+
+
+ )} +
+ )} +
+ ))} +
)} - -
{/* Input */} -
-
+ +
setInput(e.target.value)} placeholder="Ask a policy question..." disabled={isLoading} - className="flex-1 px-4 py-2 text-sm border border-[var(--color-border)] rounded-lg focus:outline-none focus:ring-2 focus:ring-[var(--color-pe-green)] disabled:opacity-50 font-mono" + className="flex-1 px-4 py-3 text-sm font-mono border border-[var(--color-border)] rounded-xl bg-white focus:outline-none focus:ring-2 focus:ring-[var(--color-pe-green)] focus:border-transparent disabled:opacity-50 placeholder:text-[var(--color-text-muted)]" />
+ +
); } diff --git a/pyproject.toml b/pyproject.toml index 175a3e5..27eb310 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -53,6 +53,10 @@ ignore = ["E501"] # Tool descriptions need to be long testpaths = ["tests"] pythonpath = ["."] asyncio_mode = "auto" +markers = [ + "integration: tests that require external services (API keys, databases)", + "slow: tests that take a long time to run", +] [dependency-groups] dev = [ diff --git a/scripts/init.py b/scripts/init.py index 587755e..cf7a04a 100644 --- a/scripts/init.py +++ b/scripts/init.py @@ -157,6 +157,7 @@ def apply_rls_policies(engine): "parameters", "parameter_values", "users", + "household_jobs", ] # Read-only tables (public can read, only service role can write) @@ -176,6 +177,7 @@ def apply_rls_policies(engine): "policies", "dynamics", "reports", + "household_jobs", ] # Read-only results tables diff --git a/src/policyengine_api/agent_sandbox.py b/src/policyengine_api/agent_sandbox.py index 01ac0fe..6408093 100644 --- a/src/policyengine_api/agent_sandbox.py +++ b/src/policyengine_api/agent_sandbox.py @@ -2,6 +2,7 @@ import json import re +import time from typing import Any, Callable import anthropic @@ -34,8 +35,27 @@ 3. Be concise but thorough 4. For UK, amounts are in GBP. For US, amounts are in USD. 5. Poll async endpoints until status is "completed" + +IMPORTANT: When polling async endpoints, ALWAYS use the sleep tool to wait 5-10 seconds between requests. +Do not poll in a tight loop - this wastes resources and may hit rate limits. """ +# Sleep tool for polling delays +SLEEP_TOOL = { + "name": "sleep", + "description": "Wait for a specified number of seconds. Use this between polling requests to avoid hammering the API.", + "input_schema": { + "type": "object", + "properties": { + "seconds": { + "type": "number", + "description": "Number of seconds to sleep (1-60)", + } + }, + "required": ["seconds"], + }, +} + def fetch_openapi_spec(api_base_url: str) -> dict: """Fetch and cache OpenAPI spec.""" @@ -285,7 +305,7 @@ def _run_agent_impl( question: str, api_base_url: str = "https://v2.api.policyengine.org", call_id: str = "", - max_turns: int = 15, + max_turns: int = 30, ) -> dict: """Core agent implementation.""" @@ -316,6 +336,8 @@ def log(msg: str) -> None: claude_tools = [ {k: v for k, v in t.items() if k != "_meta"} for t in tools ] + # Add the sleep tool + claude_tools.append(SLEEP_TOOL) client = anthropic.Anthropic() messages = [{"role": "user", "content": question}] @@ -350,11 +372,18 @@ def log(msg: str) -> None: assistant_content.append(block) # Execute tool - tool = tool_lookup.get(block.name) - if tool: - result = execute_api_tool(tool, block.input, api_base_url, log) + if block.name == "sleep": + # Handle sleep tool specially + seconds = min(max(block.input.get("seconds", 5), 1), 60) + log(f"[SLEEP] Waiting {seconds} seconds...") + time.sleep(seconds) + result = f"Slept for {seconds} seconds" else: - result = f"Unknown tool: {block.name}" + tool = tool_lookup.get(block.name) + if tool: + result = execute_api_tool(tool, block.input, api_base_url, log) + else: + result = f"Unknown tool: {block.name}" log(f"[TOOL_RESULT] {result[:300]}") @@ -392,12 +421,12 @@ def log(msg: str) -> None: return result -@app.function(image=image, secrets=[anthropic_secret], timeout=300) +@app.function(image=image, secrets=[anthropic_secret], timeout=600) def run_agent( question: str, api_base_url: str = "https://v2.api.policyengine.org", call_id: str = "", - max_turns: int = 15, + max_turns: int = 30, ) -> dict: """Run agentic loop to answer a policy question (Modal wrapper).""" return _run_agent_impl(question, api_base_url, call_id, max_turns) diff --git a/src/policyengine_api/api/agent.py b/src/policyengine_api/api/agent.py index 33a4f21..7389211 100644 --- a/src/policyengine_api/api/agent.py +++ b/src/policyengine_api/api/agent.py @@ -3,14 +3,14 @@ This endpoint lets users ask natural language questions about tax/benefit policy and get AI-generated reports using Claude Code connected to the PolicyEngine MCP server. -The agent runs in a Modal sandbox and logs are fetched via Modal SDK. +The agent runs in a Modal sandbox (production) or locally (development). """ +import asyncio import uuid from datetime import datetime import logfire -import modal from fastapi import APIRouter, HTTPException from pydantic import BaseModel @@ -67,6 +67,19 @@ class StatusResponse(BaseModel): _logs: dict[str, list[LogEntry]] = {} +def _run_local_agent(call_id: str, question: str, api_base_url: str) -> None: + """Run agent locally in a background thread.""" + from policyengine_api.agent_sandbox import _run_agent_impl + + try: + result = _run_agent_impl(question, api_base_url, call_id) + _calls[call_id]["status"] = result.get("status", "completed") + _calls[call_id]["result"] = result + except Exception as e: + _calls[call_id]["status"] = "failed" + _calls[call_id]["result"] = {"status": "failed", "error": str(e)} + + @router.post("/run", response_model=RunResponse) async def run_agent(request: RunRequest) -> RunResponse: """Start the agent to answer a policy question. @@ -90,30 +103,44 @@ async def run_agent(request: RunRequest) -> RunResponse: logfire.info("agent_run", question=request.question[:100]) api_base_url = settings.policyengine_api_url - - # Look up the deployed function - run_fn = modal.Function.from_name("policyengine-sandbox", "run_agent") - - # Generate a call_id before spawning so we can pass it to the function call_id = f"fc-{uuid.uuid4().hex[:24]}" # Initialize logs storage _logs[call_id] = [] - # Spawn the function (non-blocking) - pass call_id so it can POST logs back - call = run_fn.spawn(request.question, api_base_url, call_id) - - # Store call info - _calls[call_id] = { - "call": call, - "modal_call_id": call.object_id, - "question": request.question, - "started_at": datetime.utcnow().isoformat(), - "status": "running", - "result": None, - } - - logfire.info("agent_spawned", call_id=call_id, modal_call_id=call.object_id) + if settings.agent_use_modal: + # Production: use Modal + import modal + + run_fn = modal.Function.from_name("policyengine-sandbox", "run_agent") + call = run_fn.spawn(request.question, api_base_url, call_id) + + _calls[call_id] = { + "call": call, + "modal_call_id": call.object_id, + "question": request.question, + "started_at": datetime.utcnow().isoformat(), + "status": "running", + "result": None, + } + logfire.info("agent_spawned", call_id=call_id, modal_call_id=call.object_id) + else: + # Local development: run in background thread + _calls[call_id] = { + "call": None, + "modal_call_id": None, + "question": request.question, + "started_at": datetime.utcnow().isoformat(), + "status": "running", + "result": None, + } + logfire.info("agent_spawned_local", call_id=call_id) + + # Run in background using asyncio + loop = asyncio.get_event_loop() + loop.run_in_executor( + None, _run_local_agent, call_id, request.question, api_base_url + ) return RunResponse(call_id=call_id, status="running") diff --git a/supabase/migrations/20241121000001_rls_policies.sql b/supabase/migrations/20241121000001_rls_policies.sql deleted file mode 100644 index 7022091..0000000 --- a/supabase/migrations/20241121000001_rls_policies.sql +++ /dev/null @@ -1,157 +0,0 @@ --- Enable RLS on all application tables -ALTER TABLE datasets ENABLE ROW LEVEL SECURITY; -ALTER TABLE dataset_versions ENABLE ROW LEVEL SECURITY; -ALTER TABLE simulations ENABLE ROW LEVEL SECURITY; -ALTER TABLE policies ENABLE ROW LEVEL SECURITY; -ALTER TABLE dynamics ENABLE ROW LEVEL SECURITY; -ALTER TABLE aggregates ENABLE ROW LEVEL SECURITY; -ALTER TABLE change_aggregates ENABLE ROW LEVEL SECURITY; -ALTER TABLE tax_benefit_models ENABLE ROW LEVEL SECURITY; -ALTER TABLE tax_benefit_model_versions ENABLE ROW LEVEL SECURITY; -ALTER TABLE variables ENABLE ROW LEVEL SECURITY; -ALTER TABLE parameters ENABLE ROW LEVEL SECURITY; -ALTER TABLE parameter_values ENABLE ROW LEVEL SECURITY; - --- Service role policies (full access to everything) -DO $$ -BEGIN - -- Datasets - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'datasets' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON datasets FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Dataset versions - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'dataset_versions' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON dataset_versions FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Simulations - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'simulations' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON simulations FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Policies - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'policies' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON policies FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Dynamics - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'dynamics' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON dynamics FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Aggregates - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'aggregates' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON aggregates FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Change aggregates - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'change_aggregates' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON change_aggregates FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Tax benefit models - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'tax_benefit_models' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON tax_benefit_models FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Tax benefit model versions - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'tax_benefit_model_versions' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON tax_benefit_model_versions FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Variables - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'variables' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON variables FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Parameters - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'parameters' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON parameters FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; - - -- Parameter values - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'parameter_values' AND policyname = 'Service role full access') THEN - CREATE POLICY "Service role full access" ON parameter_values FOR ALL TO service_role USING (true) WITH CHECK (true); - END IF; -END $$; - --- Public read access for read-only tables -DO $$ -BEGIN - -- Tax benefit models (read-only for public) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'tax_benefit_models' AND policyname = 'Public read access') THEN - CREATE POLICY "Public read access" ON tax_benefit_models FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Tax benefit model versions (read-only for public) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'tax_benefit_model_versions' AND policyname = 'Public read access') THEN - CREATE POLICY "Public read access" ON tax_benefit_model_versions FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Variables (read-only for public) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'variables' AND policyname = 'Public read access') THEN - CREATE POLICY "Public read access" ON variables FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Parameters (read-only for public) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'parameters' AND policyname = 'Public read access') THEN - CREATE POLICY "Public read access" ON parameters FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Parameter values (read-only for public) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'parameter_values' AND policyname = 'Public read access') THEN - CREATE POLICY "Public read access" ON parameter_values FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Datasets (read-only for public) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'datasets' AND policyname = 'Public read access') THEN - CREATE POLICY "Public read access" ON datasets FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Dataset versions (read-only for public) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'dataset_versions' AND policyname = 'Public read access') THEN - CREATE POLICY "Public read access" ON dataset_versions FOR SELECT TO anon, authenticated USING (true); - END IF; -END $$; - --- User-created content policies -DO $$ -BEGIN - -- Simulations (users can create and read their own) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'simulations' AND policyname = 'Users can create simulations') THEN - CREATE POLICY "Users can create simulations" ON simulations FOR INSERT TO anon, authenticated WITH CHECK (true); - END IF; - - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'simulations' AND policyname = 'Users can read simulations') THEN - CREATE POLICY "Users can read simulations" ON simulations FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Policies (users can create and read their own) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'policies' AND policyname = 'Users can create policies') THEN - CREATE POLICY "Users can create policies" ON policies FOR INSERT TO anon, authenticated WITH CHECK (true); - END IF; - - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'policies' AND policyname = 'Users can read policies') THEN - CREATE POLICY "Users can read policies" ON policies FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Dynamics (users can create and read their own) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'dynamics' AND policyname = 'Users can create dynamics') THEN - CREATE POLICY "Users can create dynamics" ON dynamics FOR INSERT TO anon, authenticated WITH CHECK (true); - END IF; - - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'dynamics' AND policyname = 'Users can read dynamics') THEN - CREATE POLICY "Users can read dynamics" ON dynamics FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Aggregates (read access for all) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'aggregates' AND policyname = 'Users can read aggregates') THEN - CREATE POLICY "Users can read aggregates" ON aggregates FOR SELECT TO anon, authenticated USING (true); - END IF; - - -- Change aggregates (read access for all) - IF NOT EXISTS (SELECT 1 FROM pg_policies WHERE schemaname = 'public' AND tablename = 'change_aggregates' AND policyname = 'Users can read change aggregates') THEN - CREATE POLICY "Users can read change aggregates" ON change_aggregates FOR SELECT TO anon, authenticated USING (true); - END IF; -END $$; diff --git a/supabase/migrations/20241228000000_household_jobs.sql b/supabase/migrations/20241228000000_household_jobs.sql deleted file mode 100644 index 758ba62..0000000 --- a/supabase/migrations/20241228000000_household_jobs.sql +++ /dev/null @@ -1,42 +0,0 @@ --- Create household_jobs table for async household calculations - -CREATE TABLE IF NOT EXISTS household_jobs ( - id UUID PRIMARY KEY DEFAULT gen_random_uuid(), - tax_benefit_model_name TEXT NOT NULL, - request_data JSONB NOT NULL, - policy_id UUID REFERENCES policies(id), - dynamic_id UUID REFERENCES dynamics(id), - status TEXT NOT NULL DEFAULT 'pending', - error_message TEXT, - result JSONB, - created_at TIMESTAMPTZ NOT NULL DEFAULT now(), - started_at TIMESTAMPTZ, - completed_at TIMESTAMPTZ -); - --- Index for polling by status -CREATE INDEX IF NOT EXISTS idx_household_jobs_status ON household_jobs(status); - --- Index for looking up by id -CREATE INDEX IF NOT EXISTS idx_household_jobs_id ON household_jobs(id); - --- Enable RLS -ALTER TABLE household_jobs ENABLE ROW LEVEL SECURITY; - --- Allow public read access (jobs are not sensitive) -CREATE POLICY "Allow public read access to household_jobs" - ON household_jobs - FOR SELECT - USING (true); - --- Allow public insert (anyone can create a job) -CREATE POLICY "Allow public insert to household_jobs" - ON household_jobs - FOR INSERT - WITH CHECK (true); - --- Allow service role to update (for Modal functions) -CREATE POLICY "Allow service role to update household_jobs" - ON household_jobs - FOR UPDATE - USING (true); diff --git a/supabase/migrations/20241229000000_allow_null_user_id.sql b/supabase/migrations/20241229000000_allow_null_user_id.sql deleted file mode 100644 index dba0825..0000000 --- a/supabase/migrations/20241229000000_allow_null_user_id.sql +++ /dev/null @@ -1,2 +0,0 @@ --- Allow null user_id in reports table for anonymous API-triggered reports -ALTER TABLE reports ALTER COLUMN user_id DROP NOT NULL; diff --git a/tests/test_agent.py b/tests/test_agent.py index c917159..2c591f5 100644 --- a/tests/test_agent.py +++ b/tests/test_agent.py @@ -3,10 +3,12 @@ Tests verify that Claude Code is invoked correctly with proper MCP configuration. """ +import pytest + +pytestmark = pytest.mark.integration + import json from unittest.mock import AsyncMock, MagicMock, patch - -import pytest from fastapi.testclient import TestClient from policyengine_api.main import app diff --git a/tests/test_analysis.py b/tests/test_analysis.py index b093303..90dbe7c 100644 --- a/tests/test_analysis.py +++ b/tests/test_analysis.py @@ -5,6 +5,8 @@ """ import pytest + +pytestmark = pytest.mark.integration from fastapi.testclient import TestClient from sqlmodel import Session, select diff --git a/tests/test_household.py b/tests/test_household.py index f8e2629..8f17176 100644 --- a/tests/test_household.py +++ b/tests/test_household.py @@ -1,6 +1,9 @@ """Tests for household calculation endpoint.""" import pytest + +pytestmark = pytest.mark.integration + from fastapi.testclient import TestClient from policyengine_api.main import app diff --git a/tests/test_household_impact.py b/tests/test_household_impact.py index 1cfd469..2ed1224 100644 --- a/tests/test_household_impact.py +++ b/tests/test_household_impact.py @@ -1,6 +1,9 @@ """Tests for household impact comparison endpoint.""" import pytest + +pytestmark = pytest.mark.integration + from fastapi.testclient import TestClient from policyengine_api.main import app diff --git a/tests/test_integration.py b/tests/test_integration.py index d3f8dc1..e044cab 100644 --- a/tests/test_integration.py +++ b/tests/test_integration.py @@ -4,9 +4,11 @@ Run with: make integration-test """ -from datetime import datetime, timezone - import pytest + +pytestmark = pytest.mark.integration + +from datetime import datetime, timezone from rich.console import Console from sqlmodel import Session, create_engine, select