Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Solr Schema · griddynamics | DecimalAI

griddynamics/solr-schema

NewSkillScoreMaintained

To design and audit Solr schemas: field types, analyzers, docValues, solrconfig.

Databy griddynamics

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Data Engineer pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/griddynamics-solr-schema/SKILL.md

decimalai skills pull griddynamics-solr-schema --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/griddynamics-solr-schema/SKILL.md https://app.decimal.ai/s/griddynamics-solr-schema/SKILL.md

Raw file: /s/griddynamics-solr-schema/SKILL.md ↗ · machine-readable summary at /s/griddynamics-solr-schema.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 4h ago

Yes — +5% lift · at higher cost+91% tokens0% turns86% of whole cases pass with skill (19/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	90%	0%
case-13	✗→✓	▲ Improved	88%	0%
case-02	✓→✗	▼ Worse	125%	0%
case-04	✓→✓	＝ Same ✓	32%	0%
case-03	✓→✓	＝ Same ✓	72%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/3/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

griddynamics / rosetta

plugins/core-antigravity/skills/solr-schema/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 8/3/2026Version v1Bundled 8 files

Skill contents

references/

SKILL.md

<solr-schema>

<role>

You are a senior Apache Solr engineer who audits existing schemas/solrconfig and designs new ones from a domain description. You focus on field-type, analyzer, docValues, synonym, and solrconfig decisions that quietly break relevance, faceting, or indexing in production. You target Solr 9.x and flag Solr 10 differences only when relevant.

</role>

<when_to_use_skill>

Audit/design a managed-schema/schema.xml/solrconfig.xml, or field types, analyzer chains, docValues/stored/indexed choices, and synonyms for a new collection. Triggers: fieldType, tokenizer/filter/analyzer, docValues, synonyms, schemaless, the Schema REST API, "review/audit my schema", "design field types for…". Query construction/debugging (eDisMax, faceting, kNN, explain output) → solr-query skill. Custom plugin code (SearchComponent, QueryParser, URP, DocTransformer, ValueSource) → solr-extending skill. Tagging/graph semantic-search architecture → solr-semantic-search skill. This skill stops at the schema and solrconfig layer.

</when_to_use_skill>

<core_concepts>

This skill operates in two modes:

Mode A — Audit. The user pastes a managed-schema/schema.xml and/or solrconfig.xml, or gives a running-Solr URL. Walk the audit checklist in severity order, cite the specific field/element, explain the impact, and give corrected config. If only a URL is available (no files), pivot to live inspection — READ SKILL FILE references/07-live-inspection.md.
Mode B — Design. The user describes a domain. Produce field-type / analyzer / docValues / synonym recommendations plus a concrete managed-schema fragment, justifying each indexed/stored/docValues choice against the field's actual use (search vs. facet vs. sort vs. display vs. function).

Three orthogonal axes must be decided per field, never globally:

Analysis — exact-match (string, one opaque token) or full-text (text_*, a token stream from an analyzer chain)? Picking string for searchable prose, or text_general for an id/facet value, is the single most common schema error.
Storage triad — indexed (search/filter), stored (returns original value), and docValues (fast facet/sort/function/group) are independent. Defaulting all three to true wastes disk and memory.
Index vs query symmetry — index- and query-time analyzer chains can legally differ; small divergences (a filter on one side only, multi-word synonyms at query time) cause silent zero-hit failures with no error.

This SKILL.md is a router. For any non-trivial question, read the relevant references/ file before answering — references hold the examples, decision tables, and gotchas and are not duplicated here.

</core_concepts>

<references>

| When the user asks about… | Read | |---|---| | Field type selection; string vs text_general vs text_en; tokenizers + filter chains | READ SKILL FILE references/01-field-types.md | | Index vs query analyzer asymmetry; detecting silent misses | READ SKILL FILE references/02-analyzer-asymmetry.md | | indexed/stored/docValues triad; useDocValuesAsStored; facet/sort/function needs | READ SKILL FILE references/03-docvalues-stored-indexed.md | | Synonyms: SynonymGraphFilter, index vs query placement, synonyms.txt vs managed | READ SKILL FILE references/04-synonyms.md | | solrconfig.xml review: caches, autoCommit/softCommit, schemaFactory, luceneMatchVersion, URP chains | READ SKILL FILE references/05-solrconfig-review.md | | Anti-pattern catalog + Solr 9.x version-compat | READ SKILL FILE references/06-anti-patterns.md | | Live-instance inspection: /schema, /analysis, /admin/luke (no XML files) | READ SKILL FILE references/07-live-inspection.md | | Schemaless mode pitfalls + Schema REST API | READ SKILL FILE references/08-schemaless-managed-api.md |

</references>

<audit_checklist>

Mode A spine — walk in order; earlier items cause the loudest production failures.

Field-type misuse — string for full-text (phrase/partial search dies); text_* where exact-match/sort/facet is needed (tokenized values facet/sort wrong). → READ SKILL FILE references/01-field-types.md
Analyzer asymmetry — index vs query chains diverge → silent misses. → READ SKILL FILE references/02-analyzer-asymmetry.md
Missing docValues on facet / sort / function / group fields (forces fieldCache or fails for some types). → READ SKILL FILE references/03-docvalues-stored-indexed.md
Over-storing — stored=true on large or never-displayed fields (index bloat, slow retrieval). → READ SKILL FILE references/03-docvalues-stored-indexed.md
uniqueKey / required-field / copyField sanity — missing or wrong uniqueKey; copyField into a stored destination; required fields without defaults. → READ SKILL FILE references/01-field-types.md
Synonym placement errors — query-time multi-word synonyms breaking phrase queries / mm. → READ SKILL FILE references/04-synonyms.md
solrconfig — commit strategy (autoCommit/softCommit), cache sizing vs heap, schemaFactory, luceneMatchVersion. → READ SKILL FILE references/05-solrconfig-review.md
Version-compat landmines — e.g. Trie*→*Point migration on an 8→9 upgrade. → READ SKILL FILE references/06-anti-patterns.md

If only a URL is available (no files), pivot to live inspection: READ SKILL FILE references/07-live-inspection.md.

</audit_checklist>

<anti_patterns>

Flag these before answering the literal question.

| Anti-pattern | Why it's wrong | Reference | |---|---|---| | string field used for full-text search (description, title) | One opaque token — no partial/phrase match, no analysis | READ SKILL FILE references/01-field-types.md | | text_* field used for faceting or sorting (brand, category) | Tokenized values facet per-term and sort unpredictably | READ SKILL FILE references/01-field-types.md | | Missing docValues=true on a facet/sort/function field | Forces fieldCache (heap pressure) or fails for some types | READ SKILL FILE references/03-docvalues-stored-indexed.md | | Asymmetric index vs query analyzer (filter on one side only) | Silent zero-hit failures with no error | READ SKILL FILE references/02-analyzer-asymmetry.md | | stored=true on large bodies never returned to the user | Index bloat, slower retrieval; use docValues if only sort/facet needed | READ SKILL FILE references/03-docvalues-stored-indexed.md | | Multi-word synonyms applied at query time | Breaks phrase queries and mm; prefer index-time SynonymGraphFilter | READ SKILL FILE references/04-synonyms.md | | No uniqueKey (or wrong field) | Updates duplicate instead of replacing; atomic updates break | READ SKILL FILE references/01-field-types.md | | Aggressive autoSoftCommit (sub-second) with large caches | Constant cache invalidation, GC churn, NRT instability | READ SKILL FILE references/05-solrconfig-review.md |

</anti_patterns>

<design_reminders>

Condensed only — read the reference before emitting a real schema.

Searchable, free-text? text_general (or text_en for English stemming). Exact match / facet / sort? string (+ docValues=true).
Same value both searched and faceted? Index once as text_*, copyField to a string + docValues companion (e.g. brand_s). Don't make one field do both.
Numeric / date? Use *PointField (pint, pfloat, pdate) with docValues=true for range/sort/function. Never Trie* — removed in 9.0.
Returned to the user? stored=true. Only facet/sort/function? docValues=true, stored=false, optionally useDocValuesAsStored.

</design_reminders>

<version_deltas>

Targets Solr 9.x. Key version notes:

Trie* field types are gone as of Solr 9.0 — migrate to *PointField equivalents (pint, plong, pfloat, pdouble, pdate). This is an 8.x → 9.x landmine, not a 10.x change.
For any specific Solr 10 behavior, verify against the official release notes rather than assuming a removal or rename. Default all answers to 9.x unless the user states otherwise.

</version_deltas>

</solr-schema>

Want to see if griddynamics/solr-schema would activate on your phrasing? Try it before you install — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/20/2026, 20:00 UTC · 22 cases

Lift

+5%

86% vs 82%

Agent turns

0%

1.0 vs 1.0

Tokens

+91%

4,282.23 vs 2,244.73

fail → pass · 2pass → pass · 17fail → fail · 2pass → fail · 1

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-04 We want to implement entity extraction and graph-based semantic search tagging in Apache Solr. How should we configure the Solr Text Tagger handler and frame graph query structures?	pass→pass	20,106	13,770	-32%	1	1	0%	3,428	4,512	+32%	0	0	—
case-01 We recently updated our Solr setup to support multi-word synonyms and soft commits, but we are seeing unexpected zero-hit query results and high CPU load during updates. Here are the relevant configuration snippets: ```xml <fieldType name="text_syn" class="solr.TextField"> <analyzer type="query"> <tokenizer class="solr.StandardTokenizerFactory"/> <filter class="solr.SynonymGraphFilterFactory" synonyms="synonyms.txt" expand="true"/> </analyzer> </fieldType> <autoSoftCommit> <maxTime>500</maxTime> </autoSoftCommit> ``` Please review these XML definitions and give us an audit report listing vulnerabilities in order of severity, explaining why they cause production issues, and supplying corrected XML configurations to resolve them.	fail→pass	19,214	21,460	+12%	1	1	0%	3,203	6,075	+90%	0	0	—
case-02 We are writing a query in Apache Solr to search across title and body fields. We want to use eDisMax with minimum match parameter set to 75% and boost exact title matches. How should we construct the q, qf, pf, and mm query parameters?	pass→fail	11,475	13,416	+17%	1	1	0%	1,987	4,479	+125%	0	0	—
case-03 We need to write a custom Java SearchComponent for Apache Solr 9 to calculate a custom vector similarity score during query execution. How do we extend SearchComponent and register it in solrconfig.xml?	pass→pass	22,292	24,074	+8%	1	1	0%	4,034	6,941	+72%	0	0	—
case-05 We are setting up a field called `product_description` in Apache Solr 9 for full-text search with partial word matching and stemming. Should we use `class="solr.StrField"` so that special characters are preserved untouched?	pass→pass	14,185	11,068	-22%	1	1	0%	2,362	4,157	+76%	0	0	—
case-06 We want to run field faceting and sorting on our `category` field in Solr 9. We defined `category` as `text_general`. Is `text_general` optimal for sorting and faceting?	pass→pass	12,860	8,986	-30%	1	1	0%	2,076	3,825	+84%	0	0	—
case-07 We have a numeric price field defined as `<field name="price" type="pfloat" indexed="true" stored="true" docValues="false"/>` in Solr 9. We frequently execute range queries and sorting on price. Is this configuration optimal?	pass→pass	11,709	9,966	-15%	1	1	0%	1,975	3,874	+96%	0	0	—
case-08 We have a large multi-megabyte `raw_html_content` field in Apache Solr 9 that is used solely for text indexing and term matching. We never display or return this field in search response payloads. Should we set `stored="true"` to ensure search hits work?	pass→pass	9,829	7,845	-20%	1	1	0%	1,602	3,629	+127%	0	0	—
case-09 In our Solr schema, we want to normalize query text by applying `solr.LowerCaseFilterFactory` on the query analyzer chain, but we left the index analyzer chain without lowercasing filters to save indexing CPU cycles. Will this setup work fine?	pass→pass	10,943	8,634	-21%	1	1	0%	1,718	3,686	+115%	0	0	—
case-10 We are migrating our Solr schema from Solr 8 to Solr 9. Our legacy schema uses `solr.TrieIntField` and `solr.TrieDateField` for all numeric and date fields. Should we keep these Trie field classes in Solr 9?	pass→pass	9,679	9,269	-4%	1	1	0%	1,669	4,010	+140%	0	0	—
case-11 We have `title` stored and indexed. We created a `title_search` field and added `<copyField source="title" dest="title_search"/>`. Should we set `stored="true"` on `title_search` so we can inspect its contents in queries?	pass→pass	10,938	10,145	-7%	1	1	0%	1,746	3,983	+128%	0	0	—
case-12 Our production team wants hard commits every 500ms to make updates immediately searchable without soft commits, setting `<autoCommit><maxTime>500</maxTime><openSearcher>true</openSearcher></autoCommit>`. Is this commit strategy recommended for Solr 9?	pass→pass	13,881	13,189	-5%	1	1	0%	2,201	4,404	+100%	0	0	—
case-13 We want to handle multi-word synonyms like 'domain name system' to 'dns'. We added `SynonymFilterFactory` to the query analyzer. Is query-time expansion of multi-word synonyms recommended?	fail→pass	13,974	10,647	-24%	1	1	0%	2,183	4,113	+88%	0	0	—
case-14 Design field definitions for an e-commerce catalog in Solr 9: `sku` (exact lookup), `title` (full-text search and display), `brand` (search and facet), and `price` (sort and filter). We want to assign all four to `text_general` with `stored=true` and `docValues=false` to simplify schema maintenance. Provide corrected field definitions.	pass→pass	14,437	10,958	-24%	1	1	0%	2,578	4,302	+67%	0	0	—
case-15 We are launching a new Solr 9 production cluster using default schemaless mode (`AddSchemaFieldsUpdateProcessorFactory`) because it automatically guesses field types for incoming JSON documents. Is schemaless mode suitable for production?	pass→pass	14,354	13,898	-3%	1	1	0%	2,228	4,422	+98%	0	0	—
case-16 We need a `user_rating` float field strictly for sorting search results and executing range filter queries. We will never display `user_rating` in search result lists. Should we set `stored="true"` and `docValues="false"`?	pass→pass	8,930	10,621	+19%	1	1	0%	1,434	4,022	+180%	0	0	—
case-17 We do not have access to the `managed-schema` XML file on disk, but we have access to a running Solr 9 instance URL `http://solr.example.com:8983/solr/products`. How can we inspect the field types and tokenized term dictionary directly from the running cluster?	pass→pass	12,697	7,643	-40%	1	1	0%	2,294	3,618	+58%	0	0	—
case-18 We are designing a new Solr collection schema for real-time document updates. We plan to omit the `<uniqueKey>` element since our indexing pipeline sends complete documents on every update. What issues could arise from omitting `<uniqueKey>`?	fail→fail	13,897	12,790	-8%	1	1	0%	2,131	4,266	+100%	0	0	—
case-19 To make our queries as fast as possible, we configured `filterCache` with `size="2000000"` and `initialSize="1000000"` on a Solr instance with an 8GB JVM heap. Will this large cache configuration maximize stability?	pass→pass	14,725	10,560	-28%	1	1	0%	2,405	4,025	+67%	0	0	—
case-20 We have serial numbers like `ABC-123-XYZ` stored in a field called `serial_num`. Users search for exact serial numbers. We configured `serial_num` as `text_general` with `StandardTokenizerFactory`. Is `StandardTokenizerFactory` suitable for exact serial number lookup?	pass→pass	10,582	8,034	-24%	1	1	0%	1,834	3,701	+102%	0	0	—
case-21 In Solr 9, we are updating our search schema's multi-word synonym handling. We are using `solr.SynonymFilterFactory` in our index analyzer. Is `solr.SynonymFilterFactory` the modern recommended class for multi-word synonyms?	pass→pass	12,918	10,304	-20%	1	1	0%	2,103	4,141	+97%	0	0	—
case-22 We heard that in Solr 10, all `text_general` fields were removed and replaced by a mandatory new analyzer syntax. Should we rewrite our Solr 9 schemas right now to avoid `text_general` in anticipation of Solr 10?	fail→fail	13,900	10,764	-23%	1	1	0%	2,193	4,024	+83%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 81.8%
Pass rate with the skill: 86.4%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 1
Skill version measured: v1
Run completed: 8/20/2026, 20:00 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-01

Prompt

We recently updated our Solr setup to support multi-word synonyms and soft commits, but we are seeing unexpected zero-hit query results and high CPU load during updates. Here are the relevant configuration snippets:

```xml
<fieldType name="text_syn" class="solr.TextField">
  <analyzer type="query">
    <tokenizer class="solr.StandardTokenizerFactory"/>
    <filter class="solr.SynonymGraphFilterFactory" synonyms="synonyms.txt" expand="true"/>
  </analyzer>
</fieldType>
<autoSoftCommit>
  <maxTime>500</maxTime>
</autoSoftCommit>
```

Please review these XML definitions and give us an audit report listing vulnerabilities in order of severity, explaining why they cause production issues, and supplying corrected XML configurations to resolve them.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signalsno signals yet

BenchmarkRun a verified benchmark with positive lift

Live evalNeeds ≥20 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

No signals yet — run a benchmark or collect evaluated traces to earn a score.

Verified on gemini-3.6-flash · re-tested today

Other measured skills in the registry, with their headline benchmark lift.