Information Search - Synthreo Builder

Information Search node for Builder - query vector store knowledge bases to retrieve relevant context documents and inject them into LLM prompts for accurate RAG-powered responses.

The Information Search node searches through data from previous workflow steps to find specific values or patterns. It can perform exact lookups or apply intelligent similarity matching to locate related content. Use it for data validation, content filtering, record deduplication, and information extraction tasks.

What This Node Does

The Information Search node examines data flowing through the workflow and searches for specific values or patterns within it. It can find exact matches or use configurable similarity matching to locate related information, then passes the found results (or the absence of results) to the next node.

Key capabilities:

Search for specific values within workflow data
Find exact matches or apply similarity matching with configurable tolerance
Filter and extract relevant information automatically
Handle multiple search criteria simultaneously
Control what is passed downstream when matches are found or not found

Configuration Parameters

Mode

Field name: mode
Type: Dropdown
- Inferred - Automatically determines the best search approach based on your data
- Custom Phrase - Define specific search terms (currently unavailable)
Default: Inferred
When to change: Keep on “Inferred” for most scenarios - it automatically optimizes search performance based on the data structure.

Find Among These Data Source Properties

Field name: inKeysPool
Type: Text
Default: Empty
Description: Specifies which data fields from previous nodes to search within. Enter property names separated by commas (e.g., customerName, email, phoneNumber). Leave empty to search across all available properties.
When to change: Enter specific property names when you want to limit searches to certain fields. This improves speed and reduces false matches.

Find Values of the Following Properties

Field name: keysPool
Type: Text
Default: Empty
Description: Defines which specific data values to find. Enter the exact property names whose values you want to extract (e.g., orderTotal, customerID, productName).
When to change: Set this when you want to extract specific values from matched records rather than returning the entire matched item.

Property Wrapper

Field name: propWrapper
Type: Text
Default: Empty
Description: Adds a consistent prefix or suffix to property names during search. Use when your data follows a consistent naming pattern (e.g., a customer_ prefix for all customer-related fields).
When to change: Use when working with structured data that uses a predictable naming convention to simplify configuration.

Search Mode

Field name: searchMode
Type: Dropdown
- Equals - Finds only exact matches
- Contains - Finds partial matches within larger text
Default: Equals
Description: Determines whether the node looks for exact string matches or partial matches within longer text.
When to change: Use “Contains” when searching within long text fields like descriptions, comments, or addresses. Use “Equals” for structured identifiers like account numbers, codes, or IDs where exact matching is required.

Advanced Similarity Settings

These settings appear when Search Mode is set to “Contains”. They control how flexibly the node matches text.

Apply Unicode Decoding

Field name: decodeUni
Type: Toggle (On/Off)
Default: Off
On: Converts special characters and symbols to standard text before searching. Useful for international text, accented characters, or data imported from different systems that may encode characters differently.
Off: Searches text exactly as provided.
When to change: Enable when working with multilingual data or content imported from external systems that may use non-standard character encoding.

Similarity Match

Field name: allowSimilar
Type: Toggle (On/Off)
Default: Off
On: Finds close matches even when text is not exactly the same. The degree of difference allowed is controlled by Max. Similarity Distance.
Off: Only finds exact matches within the “Contains” search.
When to change: Enable when dealing with user-generated content, imported data, or any data that may contain typos, abbreviations, or slight variations from the search term.

Max. Similarity Distance

Field name: maxSimilarDistance
Type: Number
Default: 2
Valid range: 0 and above
Description: Controls how different two pieces of text can be while still being considered a match. Measured as edit distance (the number of single-character insertions, deletions, or substitutions needed to transform one string into the other).
Recommended values:
- 1 to 2 - Minor typos and single-character variations
- 3 to 5 - Moderate variations and abbreviations
- 6 and above - Significant differences (use with caution as false positives increase)
When to change: Lower this for stricter matching; raise it for more flexible matching. Start at 2 and increase only if legitimate matches are being missed.

Relative / in %

Field name: simDistanceInPerc
Type: Toggle (On/Off)
Default: Off
On: Calculates similarity distance as a percentage of text length rather than an absolute character count. This makes the threshold consistent regardless of how long the compared strings are.
Off: Uses absolute character count for edit distance.
When to change: Enable when working with text fields of varying lengths to maintain consistent matching standards across short and long values.

Word Based Similarity

Field name: perWordSimilarity
Type: Toggle (On/Off)
Default: Off
On: Compares individual words rather than the entire text string. A match is found when enough individual words match.
Off: Compares the complete text as a single unit.
When to change: Enable when word order might vary but the content is similar (e.g., “John Smith” vs “Smith, John”) or when matching names and addresses where word-level comparison is more meaningful.

Include Phrases Words

Field name: perWordPhrases
Type: Toggle (On/Off)
Default: Off
On: Treats multi-word phrases as single units during word-based comparison. Prevents phrases from being split into individual words.
Off: Compares each word individually.
When to change: Enable when working with company names, product titles, or other multi-word entities that should be kept together as a unit rather than matched word by word.

Select Top N Items

Field name: selectTopN
Type: Toggle (On/Off)
Default: Off
On: Returns only the best matching results up to the number specified in Top N.
Off: Returns all matches that meet the criteria.
When to change: Enable when you want to limit results to the most relevant matches and avoid overwhelming downstream nodes with a large number of results.

Top N

Field name: topN
Type: Number
Default: 1
Valid range: 1 and above
Description: The number of top matches to return when Select Top N Items is enabled.
Recommended values:
- 1 - Single best match (useful when you need exactly one result)
- 3 to 5 - Multiple good options (useful when presenting alternatives)
- 10 and above - Comprehensive results for review workflows
When to change: Set to 1 when only the single best match is needed. Increase when presenting multiple options to the user or LLM.

Result Handling

If Found

Field name: passFoundRule
Type: Dropdown
- Pass found results - Sends the complete search results (matched items with all their properties) to the next node
- Pass values - Sends only the actual data values that were found
- Pass phrases - Sends only the search terms that produced matches
Default: Pass found results
When to change: Use “Pass values” when downstream nodes only need the data itself. Use “Pass phrases” when you need to know which search terms matched (e.g., for logging or reporting).

If Not Found

Field name: passNotFoundRule
Type: Dropdown
- Pass nothing - Sends no data to the next node when no matches are found
- Pass values - Sends the original search values even when no matches are found
- Pass phrases - Sends the search terms that did not produce matches
Default: Pass nothing
When to change: Use “Pass values” or “Pass phrases” when you need to track failed searches, handle missing data scenarios gracefully, or route no-match cases to a fallback path.

Example Usage

Customer Data Deduplication

Set Find among these data source properties to email, phone, lastName
Set Search Mode to Contains
Enable Similarity Match with Max. Similarity Distance = 2
Set If Found to Pass found results
Use the results to flag potential duplicate customer records for review

Product Catalog Matching

Set Find values of the following properties to productName, description, SKU
Set Search Mode to Contains
Enable Word Based Similarity for more flexible product name matching
Enable Select Top N Items and set Top N to 3
Set If Found to Pass values to get the actual product data

Document Clause Search

Set Find among these data source properties to contractText, clauses, terms
Set Search Mode to Contains
Enable Apply Unicode Decoding for special legal characters
Set Max. Similarity Distance to 3 for flexible clause matching
Set If Found to Pass phrases to record which terms were found

Support Ticket Routing

Set Find among these data source properties to issueDescription, category
Set Search Mode to Contains
Enable Word Based Similarity to match key terms
Enable Select Top N Items with Top N = 1
Set If Found to Pass found results to get complete routing information
Connect the result to a conditional routing node to direct tickets to the appropriate team

Best Practices

Only search within necessary data properties using Find among these data source properties to improve performance.
Use Equals mode for structured identifiers (account numbers, codes, IDs) where exact matching is appropriate. Reserve Contains mode for free-text fields.
Start with a Max. Similarity Distance of 2 and increase only if legitimate matches are being missed.
Use Select Top N Items to cap the number of results and prevent downstream nodes from being overwhelmed.
Test search configurations with real data samples before deploying. Similarity settings in particular may need tuning for your specific data.
Configure the If Not Found rule to handle missing data gracefully rather than letting the workflow fail silently when no matches are found.
Use Apply Unicode Decoding consistently when your data comes from multiple source systems with different character encoding.