Module: Legion::Extensions::Apollo::Runners::Knowledge

Includes:
Helpers::Lex
Included in:
Actor::EntityWatchdog, Client
Defined in:
lib/legion/extensions/apollo/runners/knowledge.rb

Constant Summary collapse

DOMAIN_ISOLATION =
{
  'claims_optimization' => ['claims_optimization'],
  'clinical_care'       => %w[clinical_care general],
  'general'             => :all
}.freeze
CONTENT_TYPE_ALIASES =
{
  reasoning: :concept, analysis: :concept, explanation: :concept,
  text: :observation, general: :observation, note: :observation, summary: :observation,
  rule: :procedure, step: :procedure, instruction: :procedure,
  link: :association, relation: :association, connection: :association,
  inference: :association, implication: :association
}.freeze
DEFAULT_QUERY_STATUS =
[:confirmed].freeze
UNSET =
Object.new.freeze
CONFLICT_CHECK_MAX_CHARS =
4000

Instance Method Summary collapse

Instance Method Details

#deprecate_entry(entry_id:, reason:) ⇒ Object



75
76
77
78
79
80
81
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 75

def deprecate_entry(entry_id:, reason:, **)
  {
    action:   :deprecate,
    entry_id: entry_id,
    reason:   reason
  }
end

#handle_erasure_request(agent_id:) ⇒ Object



326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 326

def handle_erasure_request(agent_id:, **)
  unless defined?(Legion::Data) && Legion::Data.respond_to?(:connection) && Legion::Data.connection
    return { deleted: 0, redacted: 0, error: 'apollo_data_not_available' }
  end

  log.warn("Apollo Knowledge.handle_erasure_request agent_id=#{agent_id}")
  conn = Legion::Data.connection

  # Delete entries solely from dead agent (not confirmed by others)
  deleted = conn[:apollo_entries]
            .where(source_agent: agent_id)
            .exclude(status: 'confirmed')
            .delete

  # Redact attribution on confirmed entries (corroborated, retain knowledge)
  redacted = conn[:apollo_entries]
             .where(source_agent: agent_id, status: 'confirmed')
             .update(source_agent: 'redacted', source_provider: nil, source_channel: nil)

  { deleted: deleted, redacted: redacted, agent_id: agent_id }
    .tap { |result| log.info("Apollo Knowledge.handle_erasure_request deleted=#{result[:deleted]} redacted=#{result[:redacted]} agent_id=#{agent_id}") } # rubocop:disable Layout/LineLength
rescue Sequel::Error => e
  handle_exception(e, level: :error, operation: 'apollo.knowledge.handle_erasure_request')
  { deleted: 0, redacted: 0, error: e.message }
end

#handle_ingest(content: nil, content_type: nil, tags: [], source_agent: 'unknown', source_provider: nil, source_channel: nil, knowledge_domain: nil, submitted_by: nil, submitted_from: nil, content_hash: nil, context: {}, skip: false) ⇒ Object

rubocop:disable Metrics/ParameterLists, Layout/LineLength



83
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
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 83

def handle_ingest(content: nil, content_type: nil, tags: [], source_agent: 'unknown', source_provider: nil, source_channel: nil, knowledge_domain: nil, submitted_by: nil, submitted_from: nil, content_hash: nil, context: {}, skip: false, **) # rubocop:disable Metrics/ParameterLists, Layout/LineLength
  return { status: :skipped } if skip

  content = normalize_text_input(content)
  log.debug("Apollo Knowledge.handle_ingest content_length=#{content.length} content_type=#{content_type} tags=#{Array(tags).size} source_agent=#{source_agent} source_channel=#{source_channel || 'nil'}") # rubocop:disable Layout/LineLength
  return { success: false, error: 'content is required' } if content.strip.empty?
  return { success: false, error: 'content_type is required' } if content_type.nil?
  return { success: false, error: 'apollo_data_not_available' } unless defined?(Legion::Data::Model::ApolloEntry)

  hash = content_hash || (defined?(Helpers::Writeback) ? Helpers::Writeback.content_hash(content) : nil)
  existing = active_duplicate_for_hash(hash)
  if existing
    log.info("Apollo Knowledge.handle_ingest deduped entry_id=#{existing.id} source_agent=#{source_agent}")
    return { success: true, entry_id: existing.id, deduped: true }
  end

  embedding = embed_text(content)
  content_type_sym = content_type.to_s
   = (tags: tags, knowledge_domain: knowledge_domain, source_agent: source_agent,
                             source_provider: source_provider, source_channel: source_channel,
                             submitted_by: , submitted_from: )

  corroborated, existing_id = find_corroboration(
    embedding, content_type_sym, [:source_agent], [:source_channel]
  )

  if corroborated
    log.info("Apollo Knowledge.handle_ingest corroborated entry_id=#{existing_id} source_agent=#{[:source_agent]}")
  else
    existing_id = create_candidate_entry(
      content: content, content_type: content_type_sym, context: context,
      metadata: , content_hash: hash, embedding: embedding
    )
  end

  upsert_expertise(source_agent: [:source_agent], domain: [:domain])

  Legion::Data::Model::ApolloAccessLog.create(
    entry_id: existing_id, agent_id: [:source_agent], action: 'ingest'
  )

  contradictions = detect_contradictions(existing_id, embedding, content)
  log.debug("Apollo Knowledge.handle_ingest complete entry_id=#{existing_id} corroborated=#{corroborated} contradictions=#{contradictions.size}")

  { success: true, entry_id: existing_id, status: corroborated ? 'corroborated' : 'candidate',
    corroborated: corroborated, contradictions: contradictions }
rescue Sequel::Error => e
  handle_exception(e, level: :error, operation: 'apollo.knowledge.handle_ingest')
  { success: false, error: e.message }
end

#handle_query(query:, limit: Helpers::GraphQuery.default_query_limit, min_confidence: Helpers::GraphQuery.default_query_min_confidence, status: UNSET, tags: nil, domain: nil, agent_id: 'unknown') ⇒ Object

rubocop:disable Layout/LineLength



134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
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
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 134

def handle_query(query:, limit: Helpers::GraphQuery.default_query_limit, min_confidence: Helpers::GraphQuery.default_query_min_confidence, status: UNSET, tags: nil, domain: nil, agent_id: 'unknown', **) # rubocop:disable Layout/LineLength
  return { success: false, error: 'apollo_data_not_available' } unless defined?(Legion::Data::Model::ApolloEntry)

  query = normalize_text_input(query)
  status_defaulted = status.equal?(UNSET)
  requested_status = status_defaulted ? DEFAULT_QUERY_STATUS : status
  log.debug("Apollo Knowledge.handle_query mode=#{browse_query?(query) ? 'browse' : 'semantic'} query_length=#{query.length} limit=#{limit} statuses=#{Array(requested_status).join(',')} status_defaulted=#{status_defaulted} tags=#{Array(tags).size} domain=#{domain || 'nil'} agent_id=#{agent_id}") # rubocop:disable Layout/LineLength
  if browse_query?(query)
    return list_entries_chronologically(query: query, limit: limit, status: requested_status,
                                        status_defaulted: status_defaulted, tags: tags, domain: domain)
  end

  embedding = embed_text(query)
  sql = Helpers::GraphQuery.build_semantic_search_sql(
    limit: limit, min_confidence: min_confidence,
    statuses: Array(requested_status).map(&:to_s), tags: tags, domain: domain
  )

  db = Legion::Data::Model::ApolloEntry.db
  entries = db.fetch(sql, embedding: Sequel.lit("'[#{embedding.join(',')}]'::vector")).all

  entries = entries.reject { |e| e[:distance].respond_to?(:nan?) && e[:distance].nan? }

  entries.each do |entry|
    Legion::Data::Model::ApolloEntry.where(id: entry[:id]).update(
      access_count: Sequel.expr(:access_count) + 1,
      confidence:   Helpers::Confidence.apply_retrieval_boost(
        confidence: entry[:confidence]
      ),
      updated_at:   Time.now
    )
  end

  if entries.any?
    Legion::Data::Model::ApolloAccessLog.create(
      entry_id: entries.first&.dig(:id), agent_id: agent_id, action: 'query'
    )
  end

  formatted = entries.map do |entry|
    { id: entry[:id], content: entry[:content], content_type: entry[:content_type],
      confidence: entry[:confidence], distance: entry[:distance]&.to_f,
      tags: entry[:tags], source_agent: entry[:source_agent],
      knowledge_domain: entry[:knowledge_domain] }
  end

  log.info("Apollo Knowledge.handle_query results=#{formatted.size} mode=semantic agent_id=#{agent_id}")
  { success: true, entries: formatted, count: formatted.size }
rescue Sequel::Error => e
  handle_exception(e, level: :error, operation: 'apollo.knowledge.handle_query')
  { success: false, error: e.message }
end

#handle_traverse(entry_id:, depth: Helpers::GraphQuery.default_depth, relation_types: nil, agent_id: 'unknown') ⇒ Object



187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 187

def handle_traverse(entry_id:, depth: Helpers::GraphQuery.default_depth, relation_types: nil, agent_id: 'unknown', **)
  return { success: false, error: 'apollo_data_not_available' } unless defined?(Legion::Data::Model::ApolloEntry)

  log.debug("Apollo Knowledge.handle_traverse entry_id=#{entry_id} depth=#{depth} relation_types=#{Array(relation_types).join(',')} agent_id=#{agent_id}") # rubocop:disable Layout/LineLength
  # Whitelist relation_types to prevent SQL injection (they are string-interpolated in build_traversal_sql)
  if relation_types
    allowed = Helpers::Confidence::RELATION_TYPES
    relation_types = relation_types.select { |t| allowed.include?(t.to_s) }
  end

  sql = Helpers::GraphQuery.build_traversal_sql(depth: depth, relation_types: relation_types)
  db = Legion::Data::Model::ApolloEntry.db
  entries = db.fetch(sql, entry_id: entry_id).all

  if entries.any? && agent_id != 'unknown'
    Legion::Data::Model::ApolloAccessLog.create(
      entry_id: entry_id, agent_id: agent_id, action: 'query'
    )
  end

  formatted = entries.map do |entry|
    { id: entry[:id], content: entry[:content], content_type: entry[:content_type],
      confidence: entry[:confidence], tags: entry[:tags], source_agent: entry[:source_agent],
      depth: entry[:depth], activation: entry[:activation] }
  end

  log.info("Apollo Knowledge.handle_traverse results=#{formatted.size} entry_id=#{entry_id}")
  { success: true, entries: formatted, count: formatted.size }
rescue Sequel::Error => e
  handle_exception(e, level: :error, operation: 'apollo.knowledge.handle_traverse')
  { success: false, error: e.message }
end

#prepare_mesh_export(target_domain:, min_confidence: Helpers::Confidence.apollo_setting(:query, :mesh_export_min_confidence, default: 0.5), limit: Helpers::Confidence.apollo_setting(:query, :mesh_export_limit, default: 100)) ⇒ Object

rubocop:disable Layout/LineLength



295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 295

def prepare_mesh_export(target_domain:, min_confidence: Helpers::Confidence.apollo_setting(:query, :mesh_export_min_confidence, default: 0.5), limit: Helpers::Confidence.apollo_setting(:query, :mesh_export_limit, default: 100), **) # rubocop:disable Layout/LineLength
  unless defined?(Legion::Data) && Legion::Data.respond_to?(:connection) && Legion::Data.connection
    return { success: false, error: 'apollo_data_not_available' }
  end

  log.debug("Apollo Knowledge.prepare_mesh_export target_domain=#{target_domain} min_confidence=#{min_confidence} limit=#{limit}")
  conn = Legion::Data.connection
  allowed = allowed_domains_for(target_domain)

  dataset = conn[:apollo_entries]
            .where(status: 'confirmed')
            .where { confidence >= min_confidence }
            .limit(limit)

  dataset = dataset.where(knowledge_domain: allowed) unless allowed == :all

  entries = dataset.all

  formatted = entries.map do |entry|
    { id: entry[:id], content: entry[:content], content_type: entry[:content_type],
      confidence: entry[:confidence], knowledge_domain: entry[:knowledge_domain],
      tags: entry[:tags], source_agent: entry[:source_agent] }
  end

  { success: true, entries: formatted, count: formatted.size, target_domain: target_domain }
    .tap { |result| log.info("Apollo Knowledge.prepare_mesh_export results=#{result[:count]} target_domain=#{target_domain}") }
rescue Sequel::Error => e
  handle_exception(e, level: :error, operation: 'apollo.knowledge.prepare_mesh_export')
  { success: false, error: e.message }
end

#query_knowledge(query:, limit: Helpers::GraphQuery.default_query_limit, min_confidence: Helpers::GraphQuery.default_query_min_confidence, status: %i[confirmed candidate],, tags: nil) ⇒ Object

rubocop:disable Layout/LineLength



46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 46

def query_knowledge(query:, limit: Helpers::GraphQuery.default_query_limit, min_confidence: Helpers::GraphQuery.default_query_min_confidence, status: %i[confirmed candidate], tags: nil, **) # rubocop:disable Layout/LineLength
  log.debug("Apollo Knowledge.query_knowledge query_length=#{query.to_s.length} limit=#{limit} statuses=#{Array(status).join(',')} tags=#{Array(tags).size} data_available=#{defined?(Legion::Data::Model::ApolloEntry) ? true : false}") # rubocop:disable Layout/LineLength
  if defined?(Legion::Data::Model::ApolloEntry)
    return handle_query(query: query, limit: limit, min_confidence: min_confidence,
                        status: status, tags: tags, **)
  end

  {
    action:         :query,
    query:          query,
    limit:          limit,
    min_confidence: min_confidence,
    status:         Array(status),
    tags:           tags
  }
end

#redistribute_knowledge(agent_id:, min_confidence: Helpers::Confidence.apollo_setting(:query, :redistribute_min_confidence, default: 0.5)) ⇒ Object



220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 220

def redistribute_knowledge(agent_id:, min_confidence: Helpers::Confidence.apollo_setting(:query, :redistribute_min_confidence, default: 0.5), **)
  return { success: false, error: 'apollo_data_not_available' } unless defined?(Legion::Data::Model::ApolloEntry)

  log.debug("Apollo Knowledge.redistribute_knowledge agent_id=#{agent_id} min_confidence=#{min_confidence}")
  entries = Legion::Data::Model::ApolloEntry
            .where(source_agent: agent_id, status: 'confirmed')
            .where { confidence > min_confidence }
            .all

  return { success: true, redistributed: 0 } if entries.empty?

  store = (Legion::Extensions::Agentic::Memory::Trace.shared_store if defined?(Legion::Extensions::Agentic::Memory::Trace))

  redistributed = 0
  entries.each do |entry|
    if store
      trace = Legion::Extensions::Agentic::Memory::Trace::Helpers::Trace.new_trace(
        type:            :semantic,
        content_payload: { content: entry.content, source_agent: agent_id,
                           content_type: entry.content_type, tags: Array(entry.tags) },
        strength:        entry.confidence.to_f,
        domain_tag:      Array(entry.tags).first || 'general'
      )
      store.store(trace)
    end
    redistributed += 1
  end

  log.info("[apollo] redistributed #{redistributed} entries from departing agent=#{agent_id}")
  { success: true, redistributed: redistributed, agent_id: agent_id }
rescue Sequel::Error => e
  handle_exception(e, level: :error, operation: 'apollo.knowledge.redistribute_knowledge')
  { success: false, error: e.message }
end


63
64
65
66
67
68
69
70
71
72
73
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 63

def related_entries(entry_id:, relation_types: nil, depth: Helpers::GraphQuery.default_depth, **)
  log.debug("Apollo Knowledge.related_entries entry_id=#{entry_id} depth=#{depth} relation_types=#{Array(relation_types).join(',')} data_available=#{defined?(Legion::Data::Model::ApolloEntry) ? true : false}") # rubocop:disable Layout/LineLength
  return handle_traverse(entry_id: entry_id, depth: depth, relation_types: relation_types, **) if defined?(Legion::Data::Model::ApolloEntry)

  {
    action:         :traverse,
    entry_id:       entry_id,
    relation_types: relation_types,
    depth:          depth
  }
end

#retrieve_relevant(query: nil, limit: Helpers::Confidence.apollo_setting(:query, :retrieval_limit, default: 5), min_confidence: Helpers::GraphQuery.default_query_min_confidence, tags: nil, domain: nil, skip: false) ⇒ Object

rubocop:disable Layout/LineLength



255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 255

def retrieve_relevant(query: nil, limit: Helpers::Confidence.apollo_setting(:query, :retrieval_limit, default: 5), min_confidence: Helpers::GraphQuery.default_query_min_confidence, tags: nil, domain: nil, skip: false, **) # rubocop:disable Layout/LineLength
  return { status: :skipped } if skip

  return { success: false, error: 'apollo_data_not_available' } unless defined?(Legion::Data::Model::ApolloEntry)

  query = normalize_text_input(query)
  log.debug("Apollo Knowledge.retrieve_relevant query_length=#{query.length} limit=#{limit} min_confidence=#{min_confidence} tags=#{Array(tags).size} domain=#{domain || 'nil'}") # rubocop:disable Layout/LineLength
  return { success: true, entries: [], count: 0 } if query.nil? || query.to_s.strip.empty?

  embedding = embed_text(query)
  sql = Helpers::GraphQuery.build_semantic_search_sql(
    limit: limit, min_confidence: min_confidence,
    statuses: %w[confirmed candidate], tags: tags, domain: domain
  )

  db = Legion::Data::Model::ApolloEntry.db
  entries = db.fetch(sql, embedding: Sequel.lit("'[#{embedding.join(',')}]'::vector")).all
  entries = entries.reject { |e| e[:distance].respond_to?(:nan?) && e[:distance].nan? }

  entries.each do |entry|
    Legion::Data::Model::ApolloEntry.where(id: entry[:id]).update(
      confidence: Helpers::Confidence.apply_retrieval_boost(confidence: entry[:confidence]),
      updated_at: Time.now
    )
  end

  formatted = entries.map do |entry|
    { id: entry[:id], content: entry[:content], content_type: entry[:content_type],
      confidence: entry[:confidence], distance: entry[:distance]&.to_f,
      tags: entry[:tags], source_agent: entry[:source_agent],
      knowledge_domain: entry[:knowledge_domain] }
  end

  log.info("Apollo Knowledge.retrieve_relevant results=#{formatted.size} limit=#{limit}")
  { success: true, entries: formatted, count: formatted.size }
rescue Sequel::Error => e
  handle_exception(e, level: :error, operation: 'apollo.knowledge.retrieve_relevant')
  { success: false, error: e.message }
end

#store_knowledge(content:, content_type:, tags: [], source_agent: nil, context: {}) ⇒ Object



27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
# File 'lib/legion/extensions/apollo/runners/knowledge.rb', line 27

def store_knowledge(content:, content_type:, tags: [], source_agent: nil, context: {}, **)
  content_type = normalize_content_type(content_type)
  log.debug("Apollo Knowledge.store_knowledge content_type=#{content_type} tags=#{Array(tags).size} source_agent=#{source_agent || 'nil'} data_available=#{defined?(Legion::Data::Model::ApolloEntry) ? true : false}") # rubocop:disable Layout/LineLength

  if defined?(Legion::Data::Model::ApolloEntry)
    return handle_ingest(content: content, content_type: content_type,
                         tags: Array(tags), source_agent: source_agent, context: context, **)
  end

  {
    action:       :store,
    content:      content,
    content_type: content_type,
    tags:         Array(tags),
    source_agent: source_agent,
    context:      context
  }
end