AI-API - Summarize Context

Opensolr AI-Hints - Demo Implementation

The Opensolr AI-Hints API, is free to use as part of your Opensolr Account.
The Opensolr AI-Hints LLM generates a summary, or answers a question, from the context you provide. This is the same API that powers the AI answers on our hosted search pages.
A number of other instructions can be passed on to this API, for NER, translation, and other capabilities.

The RAG pattern: search first, then summarize

Grounded AI answers work in two steps:

1. Retrieve — run a search on your index (for best results use hybrid search via the Solr {!hybrid} parser or the embed_and_search API endpoint) and take your top 3–4 results. embed_and_search runs the platform's tuned hybrid pipeline: your index's saved Search Tuning (Control Panel → Index Settings → Search Tuning) applies automatically, and you can override any knob per request with these optional parameters: fw_title, fw_description, fw_uri, fw_text, fw_text_t, lexical_weight, vector_weight, vector_topk (10–1000), search_mode (union / keywords_required / meaning_required / intersection), quality_boost, min_score, freshness_boost, mm (flexible / balanced / strict, or raw Solr mm syntax).
2. Summarize — concatenate the title, description and text of those results and send them to ai_summary as the context parameter, together with the user's question as query.

This keeps you in full control of retrieval — your own filters, boosts and tuning decide what the LLM reads.

1. GET or POST https://api.opensolr.com/solr_manager/api/ai_summary
2. Parameters:
   1. ​email - Required - your Opensolr registration email address
   2. api_key - Required - your Opensolr api_key
   3. ​index_name - Required - must be a valid Opensolr Index that belongs to you.
   4. ​instruction - Required - the complete prompt, and the only field that shapes the answer. What you write here is exactly what the model reads, byte for byte: nothing reorders it, relabels it, appends to it or shortens it. Up to 100,000 characters. Typically your question, then the concatenated title + description + text of your top 3–4 search results under a CONTENT: label, plus any rules the answer should follow.
   5. ​query - Deprecated - still accepted so existing integrations keep working. When present it is appended to your instruction under a QUERY: label. Put the question inside instruction instead, where you decide where it sits.
   6. ​context - Deprecated - still accepted so existing integrations keep working. When present it is appended to your instruction under a CONTENT: label. Put the content inside instruction instead.

   ---

   Full Working Examples

   cURL — Step 1: retrieve with hybrid search

   curl -X POST https://api.opensolr.com/solr_manager/api/embed_and_search \
     -d "email=your@email.com" \
     -d "api_key=YOUR_API_KEY" \
     -d "index_name=your_index" \
     -d "q=What is hybrid search?" \
     -d "rows=4" \
     -d "in=all" \
     -d "fresh=no" \
     -d "search_mode=keywords_required" \
     -d "mm=strict"
   

   The last two lines are optional per-request Search Tuning overrides — without them, your index's saved Search Tuning (or the platform defaults) applies.

   cURL — Step 2: summarize the results

   # The instruction is the whole prompt: your question, your rules, and the content you
   # retrieved in step 1. Nothing is added or rewritten between here and the model.
   curl -X POST https://api.opensolr.com/solr_manager/api/ai_summary \
     -d "email=your@email.com" \
     -d "api_key=YOUR_API_KEY" \
     -d "index_name=your_index" \
     -d "instruction=Answer this query using only the content below: What is hybrid search?
   Begin with the answer itself. No preamble, no restating the query.
   Use only what the content states.
   
   CONTENT:
   Hybrid Search - Combine keyword and vector search - Hybrid search fuses BM25 keyword scores with kNN vector similarity per document..."
   

   PHP

   <?php
   $auth = ['email' => 'your@email.com', 'api_key' => 'YOUR_API_KEY', 'index_name' => 'your_index'];
   $question = 'What is hybrid search?';
   
   // Step 1: hybrid retrieval
   $ch = curl_init('https://api.opensolr.com/solr_manager/api/embed_and_search');
   curl_setopt($ch, CURLOPT_POST, true);
   curl_setopt($ch, CURLOPT_POSTFIELDS, http_build_query($auth + [
       'q' => $question, 'rows' => 4, 'in' => 'all', 'fresh' => 'no',
   ]));
   curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
   $docs = json_decode(curl_exec($ch), true)['results']['docs'] ?? [];
   curl_close($ch);
   
   // Step 2: build the context from the top results and summarize
   $context = '';
   foreach (array_slice($docs, 0, 4) as $doc) {
       $context .= ($doc['title'] ?? '') . ' - ' . ($doc['description'] ?? '') . ' - '
                 . implode(' ', array_slice(explode(' ', $doc['text'] ?? ''), 0, 1500)) . ' - ';
   }
   
   $ch = curl_init('https://api.opensolr.com/solr_manager/api/ai_summary');
   curl_setopt($ch, CURLOPT_POST, true);
   curl_setopt($ch, CURLOPT_POSTFIELDS, http_build_query($auth + [
       'instruction' => "Answer this query using only the content below: {$question}\n"
                      . "Begin with the answer itself. No preamble, no restating the query.\n"
                      . "Use only what the content states.\n\nCONTENT:\n" . $context,
   ]));
   curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
   echo trim(curl_exec($ch));
   curl_close($ch);
   

   Python

   import requests
   
   AUTH = {"email": "your@email.com", "api_key": "YOUR_API_KEY", "index_name": "your_index"}
   question = "What is hybrid search?"
   
   # Step 1: hybrid retrieval
   resp = requests.post(
       "https://api.opensolr.com/solr_manager/api/embed_and_search",
       data={**AUTH, "q": question, "rows": 4, "in": "all", "fresh": "no"},
   )
   docs = resp.json().get("results", {}).get("docs", [])
   
   # Step 2: build the context from the top results and summarize
   context = ""
   for doc in docs[:4]:
       words = " ".join(str(doc.get("text", "")).split()[:1500])
       context += f"{doc.get('title', '')} - {doc.get('description', '')} - {words} - "
   
   resp = requests.post(
       "https://api.opensolr.com/solr_manager/api/ai_summary",
       data={
           **AUTH,
           "instruction": (
               f"Answer this query using only the content below: {question}\n"
               "Begin with the answer itself. No preamble, no restating the query.\n"
               "Use only what the content states.\n\nCONTENT:\n" + context
           ),
       },
   )
   print(resp.text.strip())