initial commit of ch5 sample app for labs (to be completed during the…

… lab)
vespa-engine · Sep 26, 2024 · 3031635 · 3031635
1 parent 266576e
commit 3031635
Show file tree

Hide file tree

Showing 6 changed files with 1,216 additions and 0 deletions.
diff --git a/examples/training-artifacts/101/ch5/embedder_rrf_rag/.vespaignore b/examples/training-artifacts/101/ch5/embedder_rrf_rag/.vespaignore
@@ -0,0 +1,6 @@
+# This file excludes unnecessary files from the application package. See
+# https://docs.vespa.ai/en/reference/vespaignore.html for more information.
+.DS_Store
+.gitignore
+README.md
+ext/
diff --git a/examples/training-artifacts/101/ch5/embedder_rrf_rag/ext/vespa_feed-1K_no_embeddings.jsonl b/examples/training-artifacts/101/ch5/embedder_rrf_rag/ext/vespa_feed-1K_no_embeddings.jsonl
diff --git a/examples/training-artifacts/101/ch5/embedder_rrf_rag/remove_embeddings.py b/examples/training-artifacts/101/ch5/embedder_rrf_rag/remove_embeddings.py
@@ -0,0 +1,15 @@
+import json
+
+INPUT_FILE="ext/vespa_feed-1K.jsonl"
+OUTPUT_FILE="ext/vespa_feed-1K_no_embeddings.jsonl"
+
+# example line in the input file
+# {"put": "id:product:product::Amazon_Fashion987231", "fields": {"id": 987231, "title": "CENTER Rabbit Party Hats Christmas Hats for Adults Bunny Hat Rabbit Ears Mobile Jumping Cap Fun Hat", "category": "Amazon_Fashion", "description": "", "price": 3100, "average_rating": 4.5, "embedding": [0.008706144988536835, 0.00692428182810545, 0.05008220300078392, -0.03883737325668335, -0.019708802923560143, 0.018436191603541374, -0.0037979180924594402, -0.04086985066533089, 0.02285926789045334, 0.022953731939196587, -0.03295670449733734, 0.01725703477859497, -0.00434730714187026, 0.036510489881038666, 0.012212918139994144, 0.007266075350344181, 0.05195595696568489, -0.003304088255390525, -0.14710889756679535, -0.01000181119889021, -0.02274658903479576, 0.04104312136769295, -0.0063829622231423855, 0.008590606041252613, -0.08836144953966141, -0.046636927872896194, -0.041007012128829956, -0.016355542466044426, -0.06989160925149918, -0.16165342926979065, -0.01774388924241066, 0.061450887471437454, -0.0012792986817657948, -0.0554032139480114, 0.041376326233148575, -0.02904839813709259, -0.013520227745175362, 0.04547768458724022, -0.018820643424987793, 0.019787374883890152, -0.022068357095122337, 0.03599236533045769, -0.018523745238780975, 0.013314939104020596, 0.013508589938282967, -0.03554629161953926, -0.015464741736650467, -0.02142908424139023, -0.04987680912017822, 0.03025384433567524, 0.03577601537108421, -0.04736088961362839, -0.02799147553741932, 0.027797596529126167, -0.006664808373898268, 0.05156460031867027, -0.003613585839048028, -0.04917958006262779, 0.045042458921670914, 0.042012769728899, -0.004005409311503172, 0.032016173005104065, -0.3252218961715698, 0.007479398977011442, 0.10389310121536255, -0.03255470097064972, -0.010276103392243385, -0.07720757275819778, -0.019704870879650116, -0.03134865686297417, 0.04496806859970093, -0.03675314784049988, -0.07906031608581543, 0.0700363740324974, 0.025187470018863678, -0.04756070300936699, 0.03532344102859497, 0.022915568202733994, 0.02770422212779522, 0.005796818993985653, -0.0692094936966896, 0.036571748554706573, 0.033657461404800415, -0.010338487103581429, 0.006327977869659662, 0.008825878612697124, -0.08306542783975601, -0.018297895789146423, -0.01921265386044979, 0.029224185273051262, -0.03989935293793678, -0.0637538954615593, 0.047146979719400406, -0.009665398858487606, -0.0600600391626358, 0.000532992766238749, 0.01686021313071251, 0.01957589201629162, 0.028927208855748177, 0.26286906003952026, 0.05114250257611275, 0.01927117444574833, 0.06915844976902008, -0.06090870499610901, 0.02024815045297146, -0.011619863100349903, -0.00921551138162613, -0.037105027586221695, -0.02271914668381214, -0.03877798467874527, 0.004896149039268494, -0.024964015930891037, 0.028943587094545364, 0.05048946291208267, -0.01663758046925068, 0.033312827348709106, -0.06477384269237518, 0.04207960143685341, 0.009631498716771603, 0.054988496005535126, -0.009936699643731117, -0.02934136800467968, -0.005706857889890671, 0.02879279851913452, -0.024603044614195824, 0.06427661329507828, -0.04814275726675987, 0.033614758402109146, -0.003578218864277005, 0.06231331825256348, 0.053646259009838104, -0.027055010199546814, -0.040031302720308304, 0.008879979141056538, -0.0320250429213047, 0.0030703917145729065, 0.06213577091693878, -0.04087560251355171, 0.007726105861365795, -0.055797725915908813, -0.026008576154708862, -0.08720946311950684, 0.13266360759735107, -0.03377920761704445, 0.025685904547572136, 0.05116606876254082, 0.033553704619407654, -0.02549905888736248, -0.03497730568051338, -0.004051984753459692, -0.015362762846052647, -0.007113694678992033, 0.025945236906409264, 0.048846080899238586, 0.03277842700481415, 0.0020529567264020443, 0.026446614414453506, 0.00017094964277930558, 0.008108239620923996, 0.023480547592043877, -0.009028516709804535, 0.04641536995768547, 0.02659882791340351, 0.08923326432704926, 0.03590584546327591, 0.0032831402495503426, 0.030130065977573395, 0.0019924382213503122, 0.055002979934215546, -0.012834412045776844, 0.02451755851507187, -0.0007795696146786213, 0.030950674787163734, 0.0358232706785202, 0.03862854093313217, -0.03138338029384613, -0.05192343145608902, 0.0330805629491806, 0.0673980712890625, -0.012659487314522266, 0.0341675840318203, 0.041247542947530746, -0.04121901094913483, 0.0033833906054496765, 0.011330662295222282, -0.0012147005181759596, -0.00838866364210844, -0.012355624698102474, 0.08863667398691177, 0.07716961205005646, -0.03360581025481224, 0.09616091102361679, -0.009030207060277462, -0.01509299036115408, 0.039576008915901184, 0.01091804914176464, -0.007640980649739504, -0.0021026721224188805, 0.10281151533126831, 0.00550538394600153, -0.061592888087034225, -0.08850890398025513, 0.0019816725980490446, -0.022194867953658104, -0.02397960238158703, -0.02136954851448536, 0.061716772615909576, -0.05318624898791313, 0.03867899253964424, -0.043909505009651184, 0.022600723430514336, -0.005250596906989813, 0.012496613897383213, 0.007864202372729778, 0.013829711824655533, 0.054033905267715454, -0.05887710675597191, -0.03665275126695633, 0.10208489745855331, 0.034644223749637604, -0.01355815026909113, 0.047204114496707916, -0.08863551914691925, -0.062089066952466965, 0.0377037487924099, 0.029051538556814194, -0.0676615834236145, -0.0495169423520565, -0.0025997627526521683, -0.015497041866183281, 0.027948834002017975, 0.026585044339299202, -0.0001117005231208168, 0.015683680772781372, 0.04765588045120239, 0.04558475688099861, -0.0683441162109375, 0.09909594804048538, 0.06924088299274445, -0.020922422409057617, 0.07137149572372437, 0.06914512813091278, 0.0004654295917134732, 0.0006155205192044377, 0.05013982206583023, -0.008206715807318687, 0.012833538465201855, -0.027573293074965477, -0.020055292174220085, 0.11351202428340912, 0.06481759995222092, -0.016648096963763237, 0.07227921485900879, -0.04866008460521698, -0.06044692173600197, 0.009577332064509392, -0.12307132035493851, -0.06041284278035164, -0.09283484518527985, 0.0684569776058197, -0.05876143276691437, -0.017231300473213196, 0.05901489779353142, -0.03168278932571411, -0.03415185585618019, -0.035074397921562195, -0.03508756682276726, 0.028621479868888855, 0.004046149551868439, -0.06223214045166969, 0.029768256470561028, 0.07121822983026505, -0.0807047039270401, -0.003811284201219678, 0.017350608482956886, -0.07254483550786972, 0.006139929872006178, 0.009430719539523125, 0.004797095432877541, 0.02165779285132885, -0.03909679129719734, -0.00018235236348118633, 0.015281707048416138, 0.011002296581864357, -0.0297059565782547, -0.06312386691570282, -0.06425154954195023, 0.007476987782865763, -0.033913422375917435, 0.020751753821969032, -0.021488163620233536, -0.030984021723270416, 0.00017515744548290968, 0.03424718603491783, 0.1250126212835312, 0.040066320449113846, 0.032705262303352356, 0.01738573983311653, -0.06829874217510223, -0.005760081112384796, -0.038384370505809784, 0.0471683032810688, -0.0015863036969676614, -0.03653369098901749, -0.10192476958036423, -0.03615661710500717, 0.017195235937833786, -0.002114292001351714, -0.04694288223981857, -0.002901688916608691, 0.03571411967277527, -0.05279149115085602, 0.011374074965715408, 0.03268689662218094, 0.07431381195783615, -0.03512423112988472, 0.01947258599102497, 0.09415441006422043, -0.014486253261566162, -0.1739281862974167, 0.0061887106858193874, 0.050733618438243866, 0.027430757880210876, -0.01607842743396759, -0.022376051172614098, 0.04357682168483734, -0.013782617636024952, 0.04539542272686958, -0.03713028505444527, 0.11902302503585815, 0.06996756792068481, 0.008093839511275291, -0.0006383286672644317, 0.0562027171254158, 0.11619176715612411, -0.0226733461022377, 0.0617365762591362, -0.04312478005886078, -0.10622943192720413, -0.04948030039668083, 0.02064296416938305, 0.12321510910987854, 0.02398318611085415, 0.016510913148522377, -0.08735989034175873, -0.0568554550409317, -0.04826837778091431, -0.09825190901756287, 0.020976798608899117, 0.03494398668408394, -0.04270603507757187, 0.053369276225566864, -0.03290951997041702, -0.023407621309161186, 0.0775999054312706, 0.0036593987606465816, -0.03759728744626045, -0.02757558785378933, -0.03483259677886963, 0.09047453105449677, 0.025219948962330818, 0.0003041496966034174, -0.013868129812180996, 0.019186925143003464, 0.012566504068672657, -0.021956535056233406, 0.022478286176919937, -0.013959426432847977, 0.011458453722298145, -0.015075542032718658, -0.015454597771167755, -0.02492937631905079, 0.03184414282441139, 0.005038146860897541, 0.03293119743466377, 0.00962645374238491, 0.008445682004094124, 0.05854005366563797, 0.05876512452960014, 0.05177081376314163, -0.07067439705133438, -0.04049496725201607, -0.07568228244781494, -0.024374980479478836]}}
+
+# remove the "embedding" field from each line
+with open(INPUT_FILE, "r") as input_file:
+    with open(OUTPUT_FILE, "w") as output_file:
+        for line in input_file:
+            line_dict = json.loads(line)
+            del line_dict["fields"]["embedding"]
+            output_file.write(json.dumps(line_dict) + "\n")
diff --git a/examples/training-artifacts/101/ch5/embedder_rrf_rag/schemas/product.sd b/examples/training-artifacts/101/ch5/embedder_rrf_rag/schemas/product.sd
@@ -0,0 +1,78 @@
+schema product {
+
+  document product {
+
+    field id type int {
+      indexing: attribute | summary
+    }
+
+    field title type string {
+      indexing: index | summary
+      index: enable-bm25
+    }
+
+    field description type string {
+      indexing: index | summary
+      index: enable-bm25
+    }
+
+    field category type string {
+      indexing: attribute | summary
+      attribute: fast-search
+      match: word
+      rank: filter
+    }
+
+    field price type int {
+      indexing: attribute | summary
+    }
+
+    field average_rating type float {
+      indexing: attribute | summary
+    }
+
+  }
+
+  # TODO define your "embedding" field here, because it will be "artificially"
+  # generated from existing fields
+
+  fieldset default {
+    fields: title, description
+  }
+
+  # semantic search rank profile. Refer to it in your query by adding:
+  # "ranking.profile": "closeness"
+  rank-profile closeness {
+    inputs {
+      query(q_embedding) tensor<float>(x[384])
+    }
+    first-phase {
+      expression: closeness(field, embedding)
+    }
+  }
+
+  # TODO hybrid search rank profile
+  # make sure to refer to it in the query
+  #
+  # rank-profile rrf inherits closeness {
+  #   function best_bm25() {
+  #     expression: max(bm25(title), bm25(description))
+  #   }
+  #   global-phase {
+  #     # TODO: you have a closeness() function inherited and a best_bm25() function defined here
+  #     # now define the reciprocal rank fusion expression
+  #     rerank-count: 200
+  #   }
+  # }
+
+  document-summary minimal {
+    summary id {}
+  }
+
+  document-summary medium {
+    summary id {}
+    summary title {}
+    summary description {}
+  }
+
+}
diff --git a/examples/training-artifacts/101/ch5/embedder_rrf_rag/services.xml b/examples/training-artifacts/101/ch5/embedder_rrf_rag/services.xml
@@ -0,0 +1,70 @@
+<?xml version="1.0" encoding="utf-8" ?>
+
+<services version="1.0" xmlns:deploy="vespa" xmlns:preprocess="properties" minimum-required-vespa-version="8.338.38">
+
+    <!-- See https://docs.vespa.ai/en/reference/services-container.html -->
+    <container id="default" version="1.0">
+
+        <clients>
+            <!-- 
+              Required for mTLS to still work, the below
+              configuration is equivalent to using a single
+              security/clients.pem
+            -->
+            <client id="mtls" permissions="read,write">
+              <certificate file="security/clients.pem"/>
+            </client>
+        </clients>
+
+        <!-- See https://docs.vespa.ai/en/embedding.html#huggingface-embedder -->
+        <!-- TODO -->
+
+        <!-- Setup the client to OpenAI -->
+        <!-- TODO -->
+
+        <!--
+            <document-api> tells the container that it should accept documents for indexing. Through the
+            Document REST API you can PUT new documents, UPDATE existing documents, and DELETE documents
+            already in the cluster.
+
+            Documents sent to the Document REST API will be passed through document processors on the way
+            to the content cluster.
+
+            See:
+             - Reference: https://docs.vespa.ai/en/reference/services-container.html#document-api
+             - Operations: https://docs.vespa.ai/en/document-v1-api-guide.html
+        -->
+        <document-api/>
+
+        <!--
+            <search> tells the container to answers queries and serve results for those queries.
+            Inside the <search /> cluster you can configure chains of "searchers" -
+            Java components processing the query and/or result.
+
+            See:
+             - Reference: https://docs.vespa.ai/en/query-api.html
+             - Searchers: https://docs.vespa.ai/en/searcher-development.html
+        -->
+        <search/> <!-- TODO in here, define a search chain pointing to your OpenAI client -->
+
+        <!--
+            <nodes> specifies the nodes that should run this cluster.
+        -->
+        <nodes>
+            <node hostalias="node1" />
+        </nodes>
+
+    </container>
+
+    <!-- See https://docs.vespa.ai/en/reference/services-content.html -->
+    <content id="text" version="1.0">
+        <min-redundancy>2</min-redundancy>
+        <documents>
+            <document type="product" mode="index" />
+        </documents>
+        <nodes>
+            <node hostalias="node1" distribution-key="0" />
+        </nodes>
+    </content>
+
+</services>
diff --git a/examples/training-artifacts/101/ch5/queries.http b/examples/training-artifacts/101/ch5/queries.http
@@ -0,0 +1,47 @@
+### did we successfully feed?
+curl -H "Content-Type:application/json" https://e6384e36.c4ba1d52.z.vespa-app.cloud/search/ -d '{
+  "yql": "select * from product where true",
+  "presentation.summary": "medium"
+}'
+
+### ANN
+curl -H "Content-Type:application/json" https://e6384e36.c4ba1d52.z.vespa-app.cloud/search/ -d '{
+    "yql": "select * from product where ({targetHits:100}nearestNeighbor(embedding,q_embedding))",
+    "ranking.profile": "closeness",
+    "presentation.summary": "medium",
+    "query_text": "Mini Stencil",
+    "input.query(q_embedding)": "embed(@query_text)"
+}'
+
+
+### RRF
+curl -H "Content-Type:application/json" https://e6384e36.c4ba1d52.z.vespa-app.cloud/search/ -d '{
+    "yql": "select * from product where title contains @query_text OR ({targetHits:100}nearestNeighbor(embedding,q_embedding))",
+    "ranking.profile": "rrf",
+    "presentation.summary": "medium",
+    "query_text": "Mini Stencil",
+    "input.query(q_embedding)": "embed(@query_text)"
+}'
+
+### OpenAI command
+vespa query \
+    --timeout 60 \
+    --header="X-LLM-API-KEY:insert-api-key-here" \
+    query="what kind of mini stencils do you have for sale?" \
+    hits=5 \
+    searchChain=openai \
+    format=sse \
+    traceLevel=1
+
+### OpenAI HTTP command. The response comes back as a stream
+curl -H "Content-Type:application/json" -H "X-LLM-API-KEY:insert-api-key-here" https://e6384e36.c4ba1d52.z.vespa-app.cloud/search/ -d '{
+    "yql": "select * from product where title contains @query_text OR ({targetHits:100}nearestNeighbor(embedding,q_embedding))",
+    "ranking.profile": "rrf",
+    "presentation.summary": "medium",
+    "query_text": "Mini Stencil",
+    "input.query(q_embedding)": "embed(@query_text)",
+    "query": "what kind of mini stencils do you have for sale?",
+    "searchChain": "openai",
+    "format": "sse",
+    "traceLevel": 0
+}'