Files
dobetternorge-tools/api/deep-research.php
T
daveadmin 4cbe0a4ac4 Add Deep Research tool — agent + rank/rerank RAG
New surface at /deep-research.php where the user pastes a question or
uploads PDF/DOCX/TXT case files and a LLM-orchestrated agent researches
the Do Better Norge legal corpus from 3-5 angles, with hybrid retrieval,
cross-encoder rerank, and synthesis that emits an inline-[n]-cited
markdown brief plus a numbered sources panel.

Uploaded documents are chunked + embedded in memory only (nomic-embed-text
via LiteLLM) and searched alongside the shared corpus during the same
request — never persisted to disk, DB, or Qdrant.

Reuses ClientRagPipeline::searchAll (hybrid + rerank), dbnV6 slice
helpers, and the existing extract.php text-extraction logic via a new
dbnToolsExtractUploadedFile() helper. Also adds dbnToolsCallGpuLlm()
helper in bootstrap.php — fixes a latent bug where LegalTools.php
was already calling that name with no definition.

Search.php is unchanged.
2026-05-15 10:30:47 +02:00

68 lines
2.5 KiB
PHP

<?php
declare(strict_types=1);
require_once __DIR__ . '/../includes/bootstrap.php';
require_once __DIR__ . '/../includes/DeepResearchAgent.php';
dbnToolsRequireMethod('POST');
dbnToolsRequireAuth();
$isMultipart = stripos((string)($_SERVER['CONTENT_TYPE'] ?? ''), 'multipart/form-data') !== false;
if ($isMultipart) {
$payloadRaw = (string)($_POST['payload'] ?? '');
if ($payloadRaw === '') {
dbnToolsError('Multipart request is missing the "payload" JSON field.', 422, 'missing_payload');
}
$input = json_decode($payloadRaw, true);
if (!is_array($input)) {
dbnToolsError('Multipart "payload" field must be valid JSON.', 422, 'invalid_payload_json');
}
} else {
$input = dbnToolsJsonInput(120000);
}
$language = dbnToolsNormalizeLanguage($input['language'] ?? 'en');
dbnToolsWithTelemetry('deep_research', $language, function () use ($input, $language) {
$seedQuery = dbnToolsString($input, 'query', 4000, false);
$pastedText = dbnToolsString($input, 'paste_text', 64000, false);
$sliceInput = $input['slices'] ?? null;
$engine = (string)($input['engine'] ?? 'azure_mini');
$controls = is_array($input['controls'] ?? null) ? $input['controls'] : [];
$uploadedFiles = [];
if (!empty($_FILES['files']) && is_array($_FILES['files']['tmp_name'] ?? null)) {
$count = count($_FILES['files']['tmp_name']);
if ($count > 5) {
dbnToolsAbort('At most 5 files can be uploaded per request.', 413, 'too_many_files');
}
for ($i = 0; $i < $count; $i++) {
$file = [
'name' => $_FILES['files']['name'][$i] ?? '',
'type' => $_FILES['files']['type'][$i] ?? '',
'tmp_name' => $_FILES['files']['tmp_name'][$i] ?? '',
'error' => $_FILES['files']['error'][$i] ?? UPLOAD_ERR_NO_FILE,
'size' => $_FILES['files']['size'][$i] ?? 0,
];
$extracted = dbnToolsExtractUploadedFile($file);
$uploadedFiles[] = [
'filename' => $extracted['filename'],
'text' => $extracted['text'],
'chars' => $extracted['chars'],
'truncated' => $extracted['truncated'],
];
}
}
return (new DbnDeepResearchAgent())->run(
$seedQuery,
$pastedText,
$uploadedFiles,
is_array($sliceInput) ? $sliceInput : [],
$engine,
$language,
$controls
);
});