Module: Parse::Agent::ResultFormatter
- Extended by:
- ResultFormatter
- Included in:
- ResultFormatter
- Defined in:
- lib/parse/agent/result_formatter.rb
Overview
The ResultFormatter transforms Parse API responses into LLM-friendly formats that are easy to understand and process.
It provides consistent structure, human-readable type descriptions, and truncates large results to fit context windows.
Constant Summary collapse
- MAX_RESULTS_DISPLAY =
Maximum number of results to include in output
50- TYPE_NAMES =
Parse field type mappings for human-readable output
{ "String" => "string", "Number" => "number", "Boolean" => "boolean", "Date" => "date/time", "Object" => "object (JSON)", "Array" => "array", "GeoPoint" => "geo location", "File" => "file", "Pointer" => "pointer (reference)", "Relation" => "relation (many-to-many)", "Bytes" => "binary data", "Polygon" => "polygon (geo shape)", "ACL" => "access control list", }.freeze
Instance Method Summary collapse
-
#format_object(class_name, object, truncated_include_fields: nil) ⇒ Hash
Format a single object.
-
#format_query_results(class_name, results, limit:, skip:, where: nil, keys: nil, order: nil, include: nil, truncated_include_fields: nil) ⇒ Hash
Format query results.
-
#format_schema(schema) ⇒ Hash
Format a single schema for detailed display.
-
#format_schemas(schemas) ⇒ Hash
Format multiple schemas for display (compact summary) Returns class names grouped by type for efficient token usage.
Instance Method Details
#format_object(class_name, object, truncated_include_fields: nil) ⇒ Hash
Format a single object
212 213 214 215 216 217 218 219 220 221 222 223 224 225 |
# File 'lib/parse/agent/result_formatter.rb', line 212 def format_object(class_name, object, truncated_include_fields: nil) envelope = { class_name: class_name, object_id: object["objectId"], created_at: object["createdAt"], updated_at: object["updatedAt"], object: simplify_object(object), } if truncated_include_fields && !truncated_include_fields.empty? envelope[:truncated_include_fields] = truncated_include_fields.transform_values { || [:dropped] } end envelope end |
#format_query_results(class_name, results, limit:, skip:, where: nil, keys: nil, order: nil, include: nil, truncated_include_fields: nil) ⇒ Hash
Format query results
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 |
# File 'lib/parse/agent/result_formatter.rb', line 153 def format_query_results(class_name, results, limit:, skip:, where: nil, keys: nil, order: nil, include: nil, truncated_include_fields: nil) total = results.size truncated = total > MAX_RESULTS_DISPLAY has_more = total >= limit displayed_results = if truncated results.first(MAX_RESULTS_DISPLAY) else results end next_call = if has_more next_args = { class_name: class_name, limit: limit, skip: skip + limit, where: where, keys: keys, order: order, include: include, }.compact { tool: "query_class", arguments: next_args } end # Surface keys-on-include auto-projection metadata so the LLM # can see which joins were narrowed and re-ask with explicit # dotted paths (`keys: ["user.iconImage"]`) if it needs fields # that were dropped. Suppress the key when nothing was auto- # projected — keeps the envelope minimal for the common case. truncated_includes_payload = if truncated_include_fields && !truncated_include_fields.empty? truncated_include_fields.transform_values { || [:dropped] }.compact end { class_name: class_name, result_count: total, pagination: { limit: limit, skip: skip, has_more: has_more, }, truncated: truncated, truncated_note: truncated ? "Showing first #{MAX_RESULTS_DISPLAY} of #{total} results" : nil, truncated_include_fields: truncated_includes_payload, next_call: next_call, results: displayed_results.map { |obj| simplify_object(obj) }, }.compact end |
#format_schema(schema) ⇒ Hash
Format a single schema for detailed display
84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 |
# File 'lib/parse/agent/result_formatter.rb', line 84 def format_schema(schema) class_name = schema["className"] fields = schema["fields"] || {} indexes = schema["indexes"] || {} clp = schema["classLevelPermissions"] || {} agent_methods = schema["agent_methods"] || [] result = { class_name: class_name, type: class_type(class_name), } # Include class description if present result[:description] = schema["description"] if schema["description"] # Include analytics usage hint if present (separate from description) result[:usage] = schema["usage"] if schema["usage"] result[:fields] = format_fields_detailed(fields) result[:indexes] = format_indexes(indexes) result[:permissions] = format_clp(clp) # Include agent methods if any result[:agent_methods] = agent_methods if agent_methods.any? # Include the canonical "valid state" filter when declared. Lets # callers that opt out of the default `apply_canonical_filter` # behavior reproduce the predicate manually in their where:. if schema["canonical_filter"].is_a?(Hash) && schema["canonical_filter"].any? result[:canonical_filter] = schema["canonical_filter"] end # Echo the wire-format agent_fields allowlist when declared. The # allowlist already filters `result[:fields]` by omission, but the # explicit list answers "what may I write in `keys:` for this # class" without forcing the consumer to scan the fields array. # Storage-form columns (`_p_*`) and other Parse-internal # underscored columns are never addressable through agent tools. if schema["agent_fields"].is_a?(Array) && schema["agent_fields"].any? result[:agent_fields] = schema["agent_fields"] end # Echo the narrower join projection (wire-format) when declared. # Tells consumers "when this class is included on another class's # query, these are the fields you'll see." if schema["agent_join_fields"].is_a?(Array) && schema["agent_join_fields"].any? result[:agent_join_fields] = schema["agent_join_fields"] end # Include relationship edges if any (set by MetadataRegistry) if schema["relations"].is_a?(Hash) && (schema["relations"]["outgoing"].to_a.any? || schema["relations"]["incoming"].to_a.any?) result[:relations] = schema["relations"] end result end |
#format_schemas(schemas) ⇒ Hash
Format multiple schemas for display (compact summary) Returns class names grouped by type for efficient token usage. Use get_schema for detailed field info on specific classes.
41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 |
# File 'lib/parse/agent/result_formatter.rb', line 41 def format_schemas(schemas) built_in = [] custom = [] schemas.each do |schema| class_name = schema["className"] fields = schema["fields"] || {} agent_methods = schema["agent_methods"] || [] # Subtract the four system fields (objectId, createdAt, updatedAt, # ACL) when reporting a "user-meaningful" count, but never let the # subtraction go negative — the allowlist filter in enriched_schema # may have already trimmed system fields out. info = { name: class_name, fields: [fields.size - 4, 0].max, } # Include description if present (compact) info[:desc] = schema["description"] if schema["description"] # Include agent methods count if any info[:methods] = agent_methods.size if agent_methods.any? if class_name.start_with?("_") built_in << info else custom << info end end { total: schemas.size, note: "Use get_schema(class_name) for detailed field info", built_in: built_in, custom: custom, } end |