@echofiles/echo-pdf 0.4.3 → 0.6.0

package/dist/worker.js

@@ -1,386 +0,0 @@
-import { normalizeReturnMode } from "./file-utils.js";
-import { FileStoreDO } from "./file-store-do.js";
-import { resolveModelForProvider, resolveProviderAlias } from "./agent-defaults.js";
-import { checkHeaderAuth } from "./auth.js";
-import { handleMcpRequest } from "./mcp-server.js";
-import { loadEchoPdfConfig } from "./pdf-config.js";
-import { getRuntimeFileStore } from "./pdf-storage.js";
-import { listProviderModels } from "./provider-client.js";
-import { buildToolOutputEnvelope } from "./response-schema.js";
-import { callTool, listToolSchemas } from "./tool-registry.js";
-const json = (data, status = 200) => new Response(JSON.stringify(data), {
-    status,
-    headers: {
-        "Content-Type": "application/json; charset=utf-8",
-        "Cache-Control": "no-store",
-    },
-});
-const toError = (error) => error instanceof Error ? error.message : String(error);
-const errorStatus = (error) => {
-    const status = error?.status;
-    return typeof status === "number" && Number.isFinite(status) ? status : null;
-};
-const errorCode = (error) => {
-    const code = error?.code;
-    return typeof code === "string" && code.length > 0 ? code : null;
-};
-const errorDetails = (error) => error?.details;
-const jsonError = (error, fallbackStatus = 500) => {
-    const status = errorStatus(error) ?? fallbackStatus;
-    const code = errorCode(error);
-    const details = errorDetails(error);
-    return json({ error: toError(error), code, details }, status);
-};
-const readJson = async (request) => {
-    try {
-        const body = await request.json();
-        if (typeof body === "object" && body !== null && !Array.isArray(body)) {
-            return body;
-        }
-        return {};
-    }
-    catch {
-        return {};
-    }
-};
-const asObj = (value) => typeof value === "object" && value !== null && !Array.isArray(value)
-    ? value
-    : {};
-const resolvePublicBaseUrl = (request, configured) => typeof configured === "string" && configured.length > 0 ? configured : request.url;
-const sanitizeDownloadFilename = (filename) => {
-    const cleaned = filename
-        .replace(/[\r\n"]/g, "")
-        .replace(/[^\x20-\x7E]+/g, "")
-        .trim();
-    return cleaned.length > 0 ? cleaned : "download.bin";
-};
-const sseResponse = (stream) => new Response(stream, {
-    headers: {
-        "Content-Type": "text/event-stream; charset=utf-8",
-        "Cache-Control": "no-store",
-        Connection: "keep-alive",
-    },
-});
-const encodeSse = (event, data) => {
-    const encoder = new TextEncoder();
-    return encoder.encode(`event: ${event}\ndata: ${JSON.stringify(data)}\n\n`);
-};
-const isValidOperation = (value) => value === "extract_pages" || value === "ocr_pages" || value === "tables_to_latex";
-const toPdfOperation = (input, defaultProvider) => ({
-    operation: isValidOperation(input.operation) ? input.operation : "extract_pages",
-    fileId: typeof input.fileId === "string" ? input.fileId : undefined,
-    url: typeof input.url === "string" ? input.url : undefined,
-    base64: typeof input.base64 === "string" ? input.base64 : undefined,
-    filename: typeof input.filename === "string" ? input.filename : undefined,
-    pages: Array.isArray(input.pages) ? input.pages.map((v) => Number(v)) : [],
-    renderScale: typeof input.renderScale === "number" ? input.renderScale : undefined,
-    provider: typeof input.provider === "string" ? input.provider : defaultProvider,
-    model: typeof input.model === "string" ? input.model : "",
-    providerApiKeys: typeof input.providerApiKeys === "object" && input.providerApiKeys !== null
-        ? input.providerApiKeys
-        : undefined,
-    returnMode: normalizeReturnMode(input.returnMode),
-    prompt: typeof input.prompt === "string" ? input.prompt : undefined,
-});
-const toolNameByOperation = {
-    extract_pages: "pdf_extract_pages",
-    ocr_pages: "pdf_ocr_pages",
-    tables_to_latex: "pdf_tables_to_latex",
-};
-const operationArgsFromRequest = (request) => {
-    const args = {
-        pages: request.pages,
-    };
-    if (request.fileId)
-        args.fileId = request.fileId;
-    if (request.url)
-        args.url = request.url;
-    if (request.base64)
-        args.base64 = request.base64;
-    if (request.filename)
-        args.filename = request.filename;
-    if (typeof request.renderScale === "number")
-        args.renderScale = request.renderScale;
-    if (request.returnMode)
-        args.returnMode = request.returnMode;
-    if (request.provider)
-        args.provider = request.provider;
-    if (request.model)
-        args.model = request.model;
-    if (request.prompt)
-        args.prompt = request.prompt;
-    return args;
-};
-const checkComputeAuth = (request, env, config) => checkHeaderAuth(request, env, {
-    authHeader: config.service.computeAuth?.authHeader,
-    authEnv: config.service.computeAuth?.authEnv,
-    allowMissingSecret: false,
-    misconfiguredCode: "COMPUTE_AUTH_MISCONFIGURED",
-    unauthorizedCode: "UNAUTHORIZED",
-    contextName: "compute endpoint",
-});
-export default {
-    async fetch(request, env, ctx) {
-        const url = new URL(request.url);
-        const config = loadEchoPdfConfig(env);
-        const runtimeStore = getRuntimeFileStore(env, config);
-        const fileStore = runtimeStore.store;
-        if (request.method === "GET" && url.pathname === "/health") {
-            return json({ ok: true, service: config.service.name, now: new Date().toISOString() });
-        }
-        if (request.method === "GET" && url.pathname === "/config") {
-            return json({
-                service: config.service,
-                agent: config.agent,
-                providers: Object.entries(config.providers).map(([alias, provider]) => ({ alias, type: provider.type })),
-                capabilities: {
-                    toolCatalogEndpoint: "/tools/catalog",
-                    toolCallEndpoint: "/tools/call",
-                    fileOpsEndpoint: "/api/files/op",
-                    fileUploadEndpoint: "/api/files/upload",
-                    fileStatsEndpoint: "/api/files/stats",
-                    fileCleanupEndpoint: "/api/files/cleanup",
-                    supportedReturnModes: ["inline", "file_id", "url"],
-                },
-                mcp: {
-                    serverName: config.mcp.serverName,
-                    version: config.mcp.version,
-                    authHeader: config.mcp.authHeader ?? null,
-                },
-                fileGet: {
-                    authHeader: config.service.fileGet?.authHeader ?? null,
-                    cacheTtlSeconds: config.service.fileGet?.cacheTtlSeconds ?? 300,
-                },
-            });
-        }
-        if (request.method === "GET" && url.pathname === "/tools/catalog") {
-            return json({ tools: listToolSchemas() });
-        }
-        if (request.method === "POST" && url.pathname === "/tools/call") {
-            const auth = checkComputeAuth(request, env, config);
-            if (!auth.ok)
-                return json({ error: auth.message, code: auth.code }, auth.status);
-            const body = await readJson(request);
-            const name = typeof body.name === "string" ? body.name : "";
-            if (!name)
-                return json({ error: "Missing required field: name" }, 400);
-            try {
-                const args = asObj(body.arguments);
-                const preferredProvider = resolveProviderAlias(config, typeof body.provider === "string" ? body.provider : undefined);
-                const preferredModel = resolveModelForProvider(config, preferredProvider, typeof body.model === "string" ? body.model : undefined);
-                if (name === "pdf_ocr_pages" || name === "pdf_tables_to_latex") {
-                    if (typeof args.provider !== "string" || args.provider.length === 0) {
-                        args.provider = preferredProvider;
-                    }
-                    if (typeof args.model !== "string" || args.model.length === 0) {
-                        args.model = preferredModel;
-                    }
-                }
-                const result = await callTool(name, args, {
-                    config,
-                    env,
-                    fileStore,
-                    providerApiKeys: typeof body.providerApiKeys === "object" && body.providerApiKeys !== null
-                        ? body.providerApiKeys
-                        : undefined,
-                });
-                return json(buildToolOutputEnvelope(result, resolvePublicBaseUrl(request, config.service.publicBaseUrl)));
-            }
-            catch (error) {
-                return jsonError(error, 500);
-            }
-        }
-        if (request.method === "POST" && url.pathname === "/providers/models") {
-            const auth = checkComputeAuth(request, env, config);
-            if (!auth.ok)
-                return json({ error: auth.message, code: auth.code }, auth.status);
-            const body = await readJson(request);
-            const provider = resolveProviderAlias(config, typeof body.provider === "string" ? body.provider : undefined);
-            const runtimeKeys = typeof body.providerApiKeys === "object" && body.providerApiKeys !== null
-                ? body.providerApiKeys
-                : undefined;
-            try {
-                const models = await listProviderModels(config, env, provider, runtimeKeys);
-                return json({ provider, models });
-            }
-            catch (error) {
-                return jsonError(error, 500);
-            }
-        }
-        if (request.method === "POST" && url.pathname === "/api/agent/run") {
-            const auth = checkComputeAuth(request, env, config);
-            if (!auth.ok)
-                return json({ error: auth.message, code: auth.code }, auth.status);
-            const body = await readJson(request);
-            if (Object.hasOwn(body, "operation") && !isValidOperation(body.operation)) {
-                return json({ error: "Invalid operation. Must be one of: extract_pages, ocr_pages, tables_to_latex" }, 400);
-            }
-            const requestPayload = toPdfOperation(body, config.agent.defaultProvider);
-            try {
-                const result = await callTool(toolNameByOperation[requestPayload.operation], operationArgsFromRequest(requestPayload), {
-                    config,
-                    env,
-                    fileStore,
-                    providerApiKeys: requestPayload.providerApiKeys,
-                });
-                return json(result);
-            }
-            catch (error) {
-                return jsonError(error, 500);
-            }
-        }
-        if (request.method === "POST" && url.pathname === "/api/agent/stream") {
-            const auth = checkComputeAuth(request, env, config);
-            if (!auth.ok)
-                return json({ error: auth.message, code: auth.code }, auth.status);
-            const body = await readJson(request);
-            if (Object.hasOwn(body, "operation") && !isValidOperation(body.operation)) {
-                return json({ error: "Invalid operation. Must be one of: extract_pages, ocr_pages, tables_to_latex" }, 400);
-            }
-            const requestPayload = toPdfOperation(body, config.agent.defaultProvider);
-            const stream = new TransformStream();
-            const writer = stream.writable.getWriter();
-            let queue = Promise.resolve();
-            const send = (event, data) => {
-                queue = queue.then(() => writer.write(encodeSse(event, data))).catch(() => undefined);
-            };
-            const run = async () => {
-                try {
-                    send("meta", { kind: "meta", startedAt: new Date().toISOString(), streaming: true });
-                    send("io", { kind: "io", direction: "input", content: requestPayload });
-                    const result = await callTool(toolNameByOperation[requestPayload.operation], operationArgsFromRequest(requestPayload), {
-                        config,
-                        env,
-                        fileStore,
-                        providerApiKeys: requestPayload.providerApiKeys,
-                        trace: (event) => send("step", event),
-                    });
-                    send("io", { kind: "io", direction: "output", content: "operation completed" });
-                    send("result", { kind: "result", output: result });
-                    send("done", { ok: true });
-                }
-                catch (error) {
-                    send("error", { kind: "error", message: toError(error) });
-                    send("done", { ok: false });
-                }
-                finally {
-                    await queue;
-                    await writer.close();
-                }
-            };
-            ctx.waitUntil(run());
-            return sseResponse(stream.readable);
-        }
-        if (request.method === "POST" && url.pathname === "/api/files/op") {
-            const body = await readJson(request);
-            try {
-                const result = await callTool("file_ops", asObj(body), {
-                    config,
-                    env,
-                    fileStore,
-                });
-                return json(result);
-            }
-            catch (error) {
-                return jsonError(error, 500);
-            }
-        }
-        if (request.method === "POST" && url.pathname === "/api/files/upload") {
-            try {
-                const formData = await request.formData();
-                const file = formData.get("file");
-                if (!file || typeof file.arrayBuffer !== "function") {
-                    return json({ error: "Missing file field: file" }, 400);
-                }
-                const bytes = new Uint8Array(await file.arrayBuffer());
-                const stored = await fileStore.put({
-                    filename: file.name || `upload-${Date.now()}.pdf`,
-                    mimeType: file.type || "application/pdf",
-                    bytes,
-                });
-                return json({ file: stored }, 200);
-            }
-            catch (error) {
-                return jsonError(error, 500);
-            }
-        }
-        if (request.method === "GET" && url.pathname === "/api/files/get") {
-            const fileGetConfig = config.service.fileGet ?? {};
-            const auth = checkHeaderAuth(request, env, {
-                authHeader: fileGetConfig.authHeader,
-                authEnv: fileGetConfig.authEnv,
-                allowMissingSecret: env.ECHO_PDF_ALLOW_MISSING_AUTH_SECRET === "1",
-                misconfiguredCode: "AUTH_MISCONFIGURED",
-                unauthorizedCode: "UNAUTHORIZED",
-                contextName: "file get",
-            });
-            if (!auth.ok) {
-                return json({ error: auth.message, code: auth.code }, auth.status);
-            }
-            const fileId = url.searchParams.get("fileId") || "";
-            if (!fileId)
-                return json({ error: "Missing fileId" }, 400);
-            const file = await fileStore.get(fileId);
-            if (!file)
-                return json({ error: "File not found" }, 404);
-            const download = url.searchParams.get("download") === "1";
-            const headers = new Headers();
-            headers.set("Content-Type", file.mimeType);
-            const cacheTtl = Number(fileGetConfig.cacheTtlSeconds ?? 300);
-            const cacheControl = cacheTtl > 0
-                ? `public, max-age=${Math.floor(cacheTtl)}, s-maxage=${Math.floor(cacheTtl)}`
-                : "no-store";
-            headers.set("Cache-Control", cacheControl);
-            if (download) {
-                headers.set("Content-Disposition", `attachment; filename=\"${sanitizeDownloadFilename(file.filename)}\"`);
-            }
-            return new Response(file.bytes, { status: 200, headers });
-        }
-        if (request.method === "GET" && url.pathname === "/api/files/stats") {
-            try {
-                return json(await runtimeStore.stats(), 200);
-            }
-            catch (error) {
-                return json({ error: toError(error) }, 500);
-            }
-        }
-        if (request.method === "POST" && url.pathname === "/api/files/cleanup") {
-            try {
-                return json(await runtimeStore.cleanup(), 200);
-            }
-            catch (error) {
-                return json({ error: toError(error) }, 500);
-            }
-        }
-        if (request.method === "POST" && url.pathname === "/mcp") {
-            return await handleMcpRequest(request, env, config, fileStore);
-        }
-        if (request.method === "GET" && env.ASSETS) {
-            const assetReq = url.pathname === "/"
-                ? new Request(new URL("/index.html", url), request)
-                : request;
-            const asset = await env.ASSETS.fetch(assetReq);
-            if (asset.status !== 404)
-                return asset;
-        }
-        return json({
-            error: "Not found",
-            routes: {
-                health: "GET /health",
-                config: "GET /config",
-                toolsCatalog: "GET /tools/catalog",
-                toolCall: "POST /tools/call",
-                models: "POST /providers/models",
-                run: "POST /api/agent/run",
-                stream: "POST /api/agent/stream",
-                files: "POST /api/files/op",
-                fileUpload: "POST /api/files/upload",
-                fileGet: "GET /api/files/get?fileId=<id>",
-                fileStats: "GET /api/files/stats",
-                fileCleanup: "POST /api/files/cleanup",
-                mcp: "POST /mcp",
-            },
-        }, 404);
-    },
-};
-export { FileStoreDO };

package/scripts/export-fixtures.sh

@@ -1,204 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-ROOT_DIR="$(cd "$(dirname "$0")/.." && pwd)"
-OUT_DIR="${ROOT_DIR}/fixtures/output"
-EXPORT_PORT="${EXPORT_PORT:-8798}"
-BASE_URL="${BASE_URL:-http://127.0.0.1:${EXPORT_PORT}}"
-INPUT_PDF="${INPUT_PDF:-${ROOT_DIR}/fixtures/input.pdf}"
-START_LOCAL_DEV="${START_LOCAL_DEV:-1}"
-RUN_TABLES="${RUN_TABLES:-1}"
-REQUIRE_LLM_SUCCESS="${REQUIRE_LLM_SUCCESS:-1}"
-
-mkdir -p "$OUT_DIR"
-rm -rf "${OUT_DIR:?}/"*
-
-if [[ -f "${ROOT_DIR}/../.env.local" ]]; then
-  set -a
-  # shellcheck source=/dev/null
-  source "${ROOT_DIR}/../.env.local"
-  set +a
-fi
-
-if [[ ! -f "${INPUT_PDF}" ]]; then
-  echo "missing input pdf: ${INPUT_PDF}" >&2
-  exit 1
-fi
-
-cli() {
-  node "${ROOT_DIR}/bin/echo-pdf.js" "$@"
-}
-
-run_json() {
-  local name="$1"
-  shift
-  if "$@" > "${OUT_DIR}/${name}.json" 2> "${OUT_DIR}/${name}.err"; then
-    rm -f "${OUT_DIR}/${name}.err"
-  else
-    printf '{"ok":false,"error_file":"%s.err"}\n' "$name" > "${OUT_DIR}/${name}.json"
-  fi
-}
-
-validate_ocr_json() {
-  local json_file="$1"
-  node -e 'const fs=require("fs");const j=JSON.parse(fs.readFileSync(process.argv[1],"utf8"));const pages=j?.data?.pages;if(!Array.isArray(pages)||pages.length===0)process.exit(1);const t=String(pages[0]?.text||"").trim();if(t.length===0)process.exit(1);' "$json_file"
-}
-
-validate_tables_json() {
-  local json_file="$1"
-  node -e 'const fs=require("fs");const j=JSON.parse(fs.readFileSync(process.argv[1],"utf8"));const pages=j?.data?.pages;if(!Array.isArray(pages)||pages.length===0)process.exit(1);const t=String(pages[0]?.latex||"").trim();if(t.length===0)process.exit(1);' "$json_file"
-}
-
-# 1) Save test logs locally (do not block artifact export on transient network failure)
-set +e
-{
-  echo "[typecheck]"
-  npm --prefix "$ROOT_DIR" run typecheck
-  TYPECHECK_CODE=$?
-  echo
-  echo "[test]"
-  npm --prefix "$ROOT_DIR" run test
-  TEST_CODE=$?
-  echo
-  echo "[smoke]"
-  npm --prefix "$ROOT_DIR" run smoke
-  SMOKE_CODE=$?
-  echo
-  echo "typecheck_exit=${TYPECHECK_CODE}"
-  echo "test_exit=${TEST_CODE}"
-  echo "smoke_exit=${SMOKE_CODE}"
-} > "${OUT_DIR}/test.log" 2>&1
-set -e
-
-cat > "${OUT_DIR}/test-status.json" <<JSON
-{"typecheck":${TYPECHECK_CODE:-1},"test":${TEST_CODE:-1},"smoke":${SMOKE_CODE:-1}}
-JSON
-
-DEV_PID=""
-cleanup() {
-  if [[ -n "${DEV_PID}" ]] && kill -0 "${DEV_PID}" >/dev/null 2>&1; then
-    kill "${DEV_PID}" >/dev/null 2>&1 || true
-    wait "${DEV_PID}" 2>/dev/null || true
-  fi
-}
-trap cleanup EXIT
-
-if [[ "${START_LOCAL_DEV}" == "1" ]]; then
-  npm --prefix "$ROOT_DIR" run dev -- --ip 127.0.0.1 --port "${EXPORT_PORT}" --inspector-port 0 > "${OUT_DIR}/export-local-dev.log" 2>&1 &
-  DEV_PID=$!
-  for _ in $(seq 1 120); do
-    if node -e 'fetch(process.argv[1]+"/health").then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))' "${BASE_URL}" >/dev/null 2>&1; then
-      break
-    fi
-    sleep 0.5
-  done
-  node -e 'fetch(process.argv[1]+"/health").then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))' "${BASE_URL}" >/dev/null
-fi
-
-# 2) Init CLI + provider settings
-cli init --service-url "$BASE_URL" > "${OUT_DIR}/cli-init.json"
-
-node -e 'const fs=require("fs");const cfg=JSON.parse(fs.readFileSync(process.argv[1],"utf8"));const entries=Object.entries(cfg.providers||{});const pick=(key)=>{const keys=[key];if(key.endsWith("_API_KEY"))keys.push(key.replace(/_API_KEY$/,"_KEY"));if(key.endsWith("_KEY"))keys.push(key.replace(/_KEY$/,"_API_KEY"));for(const k of keys){const v=process.env[k];if(typeof v==="string"&&v.trim())return {k,v:v.trim()};}return null;};const forced=String(process.env.SMOKE_LLM_PROVIDER||"").trim();if(forced&&cfg.providers?.[forced]){const found=pick(String(cfg.providers[forced].apiKeyEnv||""));if(found){process.stdout.write(JSON.stringify({provider:forced,apiKey:found.v,env:found.k,forced:true}));process.exit(0);}}const preferred=String(cfg.agent?.defaultProvider||"");const ordered=entries.sort((a,b)=>a[0]===preferred?-1:b[0]===preferred?1:0);for(const [alias,p] of ordered){const found=pick(String(p.apiKeyEnv||""));if(found){process.stdout.write(JSON.stringify({provider:alias,apiKey:found.v,env:found.k,forced:false}));process.exit(0);}}process.stdout.write(JSON.stringify({provider:preferred||"",apiKey:"",env:"",forced:false}));' "${ROOT_DIR}/echo-pdf.config.json" > "${OUT_DIR}/provider-selection.json"
-PROVIDER="$(node -e 'const fs=require("fs");const j=JSON.parse(fs.readFileSync(process.argv[1],"utf8"));process.stdout.write(String(j.provider||""))' "${OUT_DIR}/provider-selection.json")"
-PROVIDER_KEY="$(node -e 'const fs=require("fs");const j=JSON.parse(fs.readFileSync(process.argv[1],"utf8"));process.stdout.write(String(j.apiKey||""))' "${OUT_DIR}/provider-selection.json")"
-PREFERRED_MODEL="${SMOKE_LLM_MODEL:-${ECHO_PDF_DEFAULT_MODEL:-}}"
-if [[ -n "${PROVIDER}" ]] && [[ -n "${PROVIDER_KEY}" ]]; then
-  cli provider set --provider "${PROVIDER}" --api-key "${PROVIDER_KEY}" > "${OUT_DIR}/provider-set.json"
-  cli provider use --provider "${PROVIDER}" > "${OUT_DIR}/provider-use.json"
-else
-  echo '{"warning":"No provider key found in env, LLM calls may fail"}' > "${OUT_DIR}/provider-warning.json"
-fi
-
-# 3) Pull models via CLI and select one
-if [[ -n "${PROVIDER}" ]]; then
-  run_json "models" cli models --provider "${PROVIDER}"
-else
-  echo '{"warning":"No provider selected, skip model list"}' > "${OUT_DIR}/models.json"
-fi
-MODEL="${PREFERRED_MODEL}"
-if [[ -n "$MODEL" ]] && [[ -n "${PROVIDER}" ]]; then
-  if ! node -e 'const fs=require("fs");const file=process.argv[1];const model=process.argv[2];const j=JSON.parse(fs.readFileSync(file,"utf8"));const models=Array.isArray(j.models)?j.models:[];process.exit(models.includes(model)?0:1)' "${OUT_DIR}/models.json" "$MODEL"; then
-    echo "Configured model not found in provider model list: ${MODEL}" >&2
-    exit 1
-  fi
-  cli model set --provider "${PROVIDER}" --model "$MODEL" > "${OUT_DIR}/model-set.json"
-else
-  echo '{"warning":"Missing ECHO_PDF_DEFAULT_MODEL / SMOKE_LLM_MODEL"}' > "${OUT_DIR}/model-warning.json"
-  exit 1
-fi
-
-# 4) Upload the exact local fixture for subsequent CLI/MCP calls
-node -e 'const fs=require("fs"); const path=require("path"); (async()=>{ const base=process.argv[1]; const file=process.argv[2]; const bytes=fs.readFileSync(file); const fd=new FormData(); fd.set("file", new Blob([bytes], {type:"application/pdf"}), path.basename(file)); const res=await fetch(`${base}/api/files/upload`, {method:"POST", body:fd}); const txt=await res.text(); fs.writeFileSync(process.argv[3], txt); if(!res.ok){process.stderr.write(txt); process.exit(1);} })().catch((e)=>{console.error(String(e)); process.exit(1)})' "$BASE_URL" "$INPUT_PDF" "${OUT_DIR}/upload.json"
-FILE_ID="$(node -e 'const fs=require("fs");const j=JSON.parse(fs.readFileSync(process.argv[1],"utf8"));process.stdout.write(j.file?.id||"")' "${OUT_DIR}/upload.json")"
-if [[ -z "${FILE_ID}" ]]; then
-  echo "upload did not return file id" >&2
-  exit 1
-fi
-
-# 5) CLI tool calls
-run_json "tools-catalog" cli tools
-if [[ -n "${PROVIDER}" ]]; then
-  run_json "cli-extract-pages" cli call --tool pdf_extract_pages --args "{\"fileId\":\"${FILE_ID}\",\"pages\":[1],\"returnMode\":\"inline\"}" --provider "${PROVIDER}" --model "${MODEL:-}"
-else
-  run_json "cli-extract-pages" cli call --tool pdf_extract_pages --args "{\"fileId\":\"${FILE_ID}\",\"pages\":[1],\"returnMode\":\"inline\"}"
-fi
-node -e 'const fs=require("fs");const p=process.argv[1];const out=process.argv[2];const j=JSON.parse(fs.readFileSync(p,"utf8"));const d=j.data?.images?.[0]?.data||"";if(!d.startsWith("data:image/"))process.exit(1);fs.writeFileSync(out, Buffer.from(d.split(",")[1]||"","base64"));' "${OUT_DIR}/cli-extract-pages.json" "${OUT_DIR}/page-1-cli.png"
-
-# 6) MCP tool calls
-run_json "mcp-initialize" cli mcp initialize
-run_json "mcp-tools" cli mcp tools
-run_json "mcp-call-fileops" cli mcp call --tool file_ops --args '{"op":"list"}'
-run_json "mcp-extract-pages" cli mcp call --tool pdf_extract_pages --args "{\"fileId\":\"${FILE_ID}\",\"pages\":[1],\"returnMode\":\"inline\"}"
-
-# 7) LLM tool calls
-OCR_OK=0
-TABLES_OK=0
-if [[ -n "${PROVIDER}" ]]; then
-  : > "${OUT_DIR}/llm-attempts.log"
-  echo "[ocr] using provider=${PROVIDER} model=${MODEL}" >> "${OUT_DIR}/llm-attempts.log"
-  if cli call --tool pdf_ocr_pages --args "{\"fileId\":\"${FILE_ID}\",\"pages\":[1],\"provider\":\"${PROVIDER}\",\"model\":\"${MODEL}\"}" --provider "${PROVIDER}" --model "${MODEL}" > "${OUT_DIR}/cli-ocr-pages.json" 2> "${OUT_DIR}/cli-ocr-pages.err"; then
-    if validate_ocr_json "${OUT_DIR}/cli-ocr-pages.json"; then
-      OCR_OK=1
-      echo "{\"provider\":\"${PROVIDER}\",\"model\":\"${MODEL}\"}" > "${OUT_DIR}/ocr-selected-model.json"
-    fi
-  fi
-else
-  run_json "cli-ocr-pages" cli call --tool pdf_ocr_pages --args "{\"fileId\":\"${FILE_ID}\",\"pages\":[1]}"
-fi
-
-if [[ "${RUN_TABLES}" == "1" ]]; then
-  if [[ -n "${PROVIDER}" ]]; then
-    echo "[tables] using provider=${PROVIDER} model=${MODEL}" >> "${OUT_DIR}/llm-attempts.log"
-    if cli call --tool pdf_tables_to_latex --args "{\"fileId\":\"${FILE_ID}\",\"pages\":[1],\"provider\":\"${PROVIDER}\",\"model\":\"${MODEL}\"}" --provider "${PROVIDER}" --model "${MODEL}" > "${OUT_DIR}/cli-tables-to-latex.json" 2> "${OUT_DIR}/cli-tables-to-latex.err"; then
-      if validate_tables_json "${OUT_DIR}/cli-tables-to-latex.json"; then
-        TABLES_OK=1
-        echo "{\"provider\":\"${PROVIDER}\",\"model\":\"${MODEL}\"}" > "${OUT_DIR}/tables-selected-model.json"
-      fi
-    fi
-  else
-    run_json "cli-tables-to-latex" cli call --tool pdf_tables_to_latex --args "{\"fileId\":\"${FILE_ID}\",\"pages\":[1]}"
-  fi
-else
-  echo '{"skipped":true,"reason":"Set RUN_TABLES=1 to enable table-latex call"}' > "${OUT_DIR}/cli-tables-to-latex.json"
-fi
-
-if [[ "${REQUIRE_LLM_SUCCESS}" == "1" ]]; then
-  if [[ "${OCR_OK}" != "1" ]]; then
-    echo "OCR failed for configured model. See ${OUT_DIR}/cli-ocr-pages.err and llm-attempts.log" >&2
-    exit 1
-  fi
-  if [[ "${RUN_TABLES}" == "1" ]] && [[ "${TABLES_OK}" != "1" ]]; then
-    echo "Tables failed for configured model. See ${OUT_DIR}/cli-tables-to-latex.err and llm-attempts.log" >&2
-    exit 1
-  fi
-fi
-
-cat > "${OUT_DIR}/summary.txt" <<TXT
-base_url=${BASE_URL}
-input_pdf=${INPUT_PDF}
-file_id=${FILE_ID}
-model=${MODEL}
-outputs_dir=${OUT_DIR}
-TXT
-
-ls -la "$OUT_DIR"

package/wrangler.toml

@@ -1,19 +0,0 @@
-name = "echo-pdf"
-main = "src/worker.ts"
-compatibility_date = "2026-03-06"
-
-[assets]
-directory = "./assets"
-binding = "ASSETS"
-
-[[r2_buckets]]
-binding = "FILE_STORE_BUCKET"
-bucket_name = "echo-pdf-files"
-
-[[durable_objects.bindings]]
-name = "FILE_STORE_DO"
-class_name = "FileStoreDO"
-
-[[migrations]]
-tag = "v1"
-new_sqlite_classes = ["FileStoreDO"]