new feature: reasoning_content for compatible models in ainvoke_response()

Tuning-Performance/Deterministic_Transactional_Workflow/agent_template_backend/.env.example

@@ -0,0 +1,211 @@
+###############################################################################
+# AI AGENT PLATFORM - CONFIGURAÇÃO ÚNICA
+# Este arquivo é lido por Pydantic Settings no framework e no backend template.
+###############################################################################
+
+APP_NAME=ai-agent-template
+APP_ENV=local
+LOG_LEVEL=INFO
+API_HOST=0.0.0.0
+API_PORT=8000
+CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173
+
+###############################################################################
+# LLM - OCI Generative AI como provider principal
+###############################################################################
+# Opções: mock, oci_openai, oci_sdk, openai_compatible
+LLM_PROVIDER=oci_openai
+LLM_TEMPERATURE=0.2
+LLM_MAX_TOKENS=2048
+LLM_TIMEOUT_SECONDS=120
+
+# OCI OpenAI-compatible endpoint
+OCI_GENAI_BASE_URL=https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1
+OCI_GENAI_MODEL=openai.gpt-4.1
+OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
+OCI_GENAI_PROJECT_OCID=
+
+# OCI_AUTH_MODE=config_file|instance_principal|resource_principal
+OCI_AUTH_MODE=config_file
+# OCI SDK / signer / profiles
+OCI_CONFIG_FILE=~/.oci/config
+OCI_PROFILE=DEFAULT
+OCI_COMPARTMENT_ID=ocid1.compartment.oc1..aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
+OCI_REGION=us-chicago-1
+
+###############################################################################
+# Persistência
+###############################################################################
+# Opções: memory, autonomous, mongodb
+SESSION_REPOSITORY_PROVIDER=autonomous
+MEMORY_REPOSITORY_PROVIDER=autonomous
+CHECKPOINT_REPOSITORY_PROVIDER=autonomous
+
+# Autonomous Database
+ADB_USER=admin
+ADB_PASSWORD=fjhsdf04954hf
+ADB_DSN=oradb23aidev_high
+ADB_WALLET_LOCATION=/ORACLE/DEFAULT/Wallet_ORADB23aiDev
+ADB_WALLET_PASSWORD=fjhsdf04954hf
+ADB_TABLE_PREFIX=AGENTFW
+
+# MongoDB - também pode representar Autonomous usando API compatível com Mongo, se habilitada no ambiente
+MONGODB_URI=mongodb://mongo:mongopassword@localhost:27017
+MONGODB_DATABASE=agent_platform
+
+# Redis
+REDIS_URL=redis://localhost:6379/0
+ENABLE_REDIS_CACHE=false
+
+###############################################################################
+# RAG / Vector / Graph
+###############################################################################
+VECTOR_STORE_PROVIDER=memory
+GRAPH_STORE_PROVIDER=memory
+RAG_TOP_K=5
+EMBEDDING_PROVIDER=mock
+OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
+RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
+
+###############################################################################
+# Observabilidade
+###############################################################################
+ENABLE_LANGFUSE=true
+LANGFUSE_TRACE_MODE=verbose # Opcional: verbose, compact
+LANGFUSE_ROOT_SPAN_NAME=agent.gateway_message
+LANGFUSE_LEGACY_IO_FALLBACK=true
+LANGFUSE_PUBLIC_KEY=pk-lf-bd9b0c7e-2b8b-4e5b-a382-284a9b4413b3
+LANGFUSE_SECRET_KEY=sk-lf-5f5cc18d-0bb5-424e-b5d0-cb3664d58c20
+LANGFUSE_HOST=http://localhost:3005
+ENABLE_OTEL=false
+OTEL_EXPORTER_OTLP_ENDPOINT=
+OTEL_SERVICE_NAME=ai-agent-template
+ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
+
+###############################################################################
+# Analytics / Observer corporativo
+###############################################################################
+# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
+ENABLE_ANALYTICS=false
+# Providers aceitos: oci_streaming,pubsub,noop
+ANALYTICS_PROVIDERS=pubsub
+# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
+AGENT_PUBSUB_TOPIC=
+GCP_PUBSUB_TOPIC_PATH=
+GCP_PROJECT_ID=
+GCP_PUBSUB_TOPIC=
+GCP_PUBSUB_TIMEOUT_SECONDS=30
+# Credencial GCP segue padrão Google:
+# GOOGLE_APPLICATION_CREDENTIALS=/secrets/gcp-service-account.json
+
+###############################################################################
+# OCI Streaming
+###############################################################################
+ENABLE_OCI_STREAMING=false
+OCI_STREAM_ENDPOINT=
+OCI_STREAM_OCID=
+OCI_STREAM_PARTITION_KEY=agent-events
+
+###############################################################################
+# Guardrails, Judges, Supervisor
+###############################################################################
+ENABLE_INPUT_GUARDRAILS=true
+ENABLE_OUTPUT_GUARDRAILS=true
+ENABLE_JUDGES=true
+ENABLE_SUPERVISOR=true
+ENABLE_OUTPUT_SUPERVISOR=true
+ENABLE_PARALLEL_GUARDRAILS=true
+GUARDRAILS_FAIL_FAST=true
+OUTPUT_SUPERVISOR_MAX_RETRIES=3
+GUARDRAILS_CONFIG_PATH=./config/guardrails.yaml
+JUDGES_CONFIG_PATH=./config/judges.yaml
+PROMPT_POLICY_PATH=./config/prompt_policy.yaml
+
+###############################################################################
+# Gateway de canais
+###############################################################################
+DEFAULT_CHANNEL=web
+# embedded = backend may parse simple/native channel payloads.
+# external = backend only accepts GatewayRequest normalized by an external Channel Gateway.
+FRAMEWORK_CHANNEL_INPUT_MODE=embedded
+ENABLE_VOICE_ADAPTER=true
+ENABLE_WHATSAPP_ADAPTER=true
+ENABLE_TEXT_ADAPTER=true
+
+#################################################
+# ENTERPRISE ROUTING
+#################################################
+# Arquivo YAML com intents, keywords, políticas de estado e fallback.
+ROUTING_CONFIG_PATH=./config/routing.yaml
+# true = usa LLM para classificar quando keywords/estado não resolverem.
+# Em produção, costuma ser útil; em desenvolvimento, false evita custo e latência.
+ENABLE_LLM_ROUTER=true
+
+# Semantic route stickiness (optional).
+# Uses a lightweight LLM profile to decide only CONTINUE vs ROUTE.
+# There are no regexes or deterministic language rules.
+ENABLE_ROUTE_STICKINESS=true
+ROUTE_STICKINESS_LLM_PROFILE=route_continuity
+ROUTE_STICKINESS_CONFIDENCE_THRESHOLD=0.90
+ROUTE_STICKINESS_HISTORY_TURNS=2
+ROUTE_STICKINESS_MAX_TOKENS=80
+HUMAN_HANDOFF_MESSAGE=Vou encaminhar seu atendimento para uma pessoa.
+END_SESSION_MESSAGE=Atendimento encerrado. Obrigado pelo contato.
+ENABLE_TRANSACTIONAL_WORKFLOWS=true
+WORKFLOWS_PATH=./workflows
+
+###############################################################################
+# MCP / Tools
+###############################################################################
+ENABLE_MCP_TOOLS=true
+MCP_SERVERS_CONFIG_PATH=./config/mcp_servers.yaml
+TOOLS_CONFIG_PATH=./config/tools.yaml
+TOOL_POLICIES_PATH=./config/tool_policies.yaml
+MCP_TOOL_TIMEOUT_SECONDS=30
+
+# router = EnterpriseRouter seleciona um agente; supervisor = pode acionar múltiplos agentes
+ROUTING_MODE=router
+
+# Usage/cost accounting
+USAGE_REPOSITORY_PROVIDER=autonomous
+IDENTITY_CONFIG_PATH=./config/identity.yaml
+MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml
+
+# -----------------------------------------------------------------------------
+# ConversationSummaryMemory / compressão de contexto conversacional
+# -----------------------------------------------------------------------------
+ENABLE_CONVERSATION_SUMMARY_MEMORY=true
+MEMORY_CONTEXT_STRATEGY=summary
+MEMORY_HISTORY_LIMIT=80
+MEMORY_RECENT_MESSAGES_LIMIT=8
+MEMORY_SUMMARY_TRIGGER_MESSAGES=20
+MEMORY_MAX_SUMMARY_CHARS=6000
+MEMORY_SUMMARY_USE_LLM=true
+MEMORY_INJECT_RECENT_MESSAGES=true
+MEMORY_INJECT_SUMMARY=true
+
+###############################################################################
+# MCP Gateway
+###############################################################################
+# true = framework routes tool calls to the dedicated MCP Gateway.
+# false = framework calls MCP servers directly from mcp_servers.yaml.
+MCP_GATEWAY_ENABLED=true
+MCP_GATEWAY_URL=http://localhost:8300
+MCP_GATEWAY_TIMEOUT_SECONDS=60
+# MCP_GATEWAY_TOKEN=
+MCP_GATEWAY_AGENT_ID=telecom_contas
+MCP_GATEWAY_TENANT_ID=default
+
+###############################################################################
+# LONG-TERM MEMORY
+###############################################################################
+ENABLE_LONG_TERM_MEMORY=true
+LONG_TERM_MEMORY_PROVIDER=sqlite
+LONG_TERM_MEMORY_SQLITE_PATH=./data/agent_framework.db
+LONG_TERM_MEMORY_TABLE=agentfw_long_term_memory
+# For Autonomous/Oracle, defaults to ${ADB_TABLE_PREFIX}_LONG_TERM_MEMORY
+# LONG_TERM_MEMORY_ORACLE_TABLE=AGENTFW_LONG_TERM_MEMORY
+LONG_TERM_MEMORY_MAX_CONTEXT_ITEMS=20
+LONG_TERM_MEMORY_MIN_CONFIDENCE=0.70
+LONG_TERM_MEMORY_AUTO_EXTRACT=true
+LONG_TERM_MEMORY_INJECT_CONTEXT=true

Tuning-Performance/Long_Term_Memory/templates/agent_template_backend/.env.example

@@ -0,0 +1,209 @@
+###############################################################################
+# AI AGENT PLATFORM - CONFIGURAÇÃO ÚNICA
+# Este arquivo é lido por Pydantic Settings no framework e no backend template.
+###############################################################################
+
+APP_NAME=ai-agent-template
+APP_ENV=local
+LOG_LEVEL=INFO
+API_HOST=0.0.0.0
+API_PORT=8000
+CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173
+
+###############################################################################
+# LLM - OCI Generative AI como provider principal
+###############################################################################
+# Opções: mock, oci_openai, oci_sdk, openai_compatible
+LLM_PROVIDER=oci_openai
+LLM_TEMPERATURE=0.2
+LLM_MAX_TOKENS=2048
+LLM_TIMEOUT_SECONDS=120
+
+# OCI OpenAI-compatible endpoint
+OCI_GENAI_BASE_URL=https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1
+OCI_GENAI_MODEL=openai.gpt-4.1
+OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
+OCI_GENAI_PROJECT_OCID=
+
+# OCI_AUTH_MODE=config_file|instance_principal|resource_principal
+OCI_AUTH_MODE=config_file
+# OCI SDK / signer / profiles
+OCI_CONFIG_FILE=~/.oci/config
+OCI_PROFILE=DEFAULT
+OCI_COMPARTMENT_ID=ocid1.compartment.oc1..aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
+OCI_REGION=us-chicago-1
+
+###############################################################################
+# Persistência
+###############################################################################
+# Opções: memory, autonomous, mongodb
+SESSION_REPOSITORY_PROVIDER=autonomous
+MEMORY_REPOSITORY_PROVIDER=autonomous
+CHECKPOINT_REPOSITORY_PROVIDER=autonomous
+
+# Autonomous Database
+ADB_USER=admin
+ADB_PASSWORD=fjhsdf04954hf
+ADB_DSN=oradb23aidev_high
+ADB_WALLET_LOCATION=/ORACLE/DEFAULT/Wallet_ORADB23aiDev
+ADB_WALLET_PASSWORD=fjhsdf04954hf
+ADB_TABLE_PREFIX=AGENTFW
+
+# MongoDB - também pode representar Autonomous usando API compatível com Mongo, se habilitada no ambiente
+MONGODB_URI=mongodb://mongo:mongopassword@localhost:27017
+MONGODB_DATABASE=agent_platform
+
+# Redis
+REDIS_URL=redis://localhost:6379/0
+ENABLE_REDIS_CACHE=false
+
+###############################################################################
+# RAG / Vector / Graph
+###############################################################################
+VECTOR_STORE_PROVIDER=memory
+GRAPH_STORE_PROVIDER=memory
+RAG_TOP_K=5
+EMBEDDING_PROVIDER=mock
+OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
+RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
+
+###############################################################################
+# Observabilidade
+###############################################################################
+ENABLE_LANGFUSE=true
+LANGFUSE_TRACE_MODE=verbose # Opcional: verbose, compact
+LANGFUSE_ROOT_SPAN_NAME=agent.gateway_message
+LANGFUSE_LEGACY_IO_FALLBACK=true
+LANGFUSE_PUBLIC_KEY=pk-lf-bd9b0c7e-2b8b-4e5b-a382-284a9b4413b3
+LANGFUSE_SECRET_KEY=sk-lf-5f5cc18d-0bb5-424e-b5d0-cb3664d58c20
+LANGFUSE_HOST=http://localhost:3005
+ENABLE_OTEL=false
+OTEL_EXPORTER_OTLP_ENDPOINT=
+OTEL_SERVICE_NAME=ai-agent-template
+ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
+
+###############################################################################
+# Analytics / Observer corporativo
+###############################################################################
+# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
+ENABLE_ANALYTICS=false
+# Providers aceitos: oci_streaming,pubsub,noop
+ANALYTICS_PROVIDERS=pubsub
+# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
+AGENT_PUBSUB_TOPIC=
+GCP_PUBSUB_TOPIC_PATH=
+GCP_PROJECT_ID=
+GCP_PUBSUB_TOPIC=
+GCP_PUBSUB_TIMEOUT_SECONDS=30
+# Credencial GCP segue padrão Google:
+# GOOGLE_APPLICATION_CREDENTIALS=/secrets/gcp-service-account.json
+
+###############################################################################
+# OCI Streaming
+###############################################################################
+ENABLE_OCI_STREAMING=false
+OCI_STREAM_ENDPOINT=
+OCI_STREAM_OCID=
+OCI_STREAM_PARTITION_KEY=agent-events
+
+###############################################################################
+# Guardrails, Judges, Supervisor
+###############################################################################
+ENABLE_INPUT_GUARDRAILS=true
+ENABLE_OUTPUT_GUARDRAILS=true
+ENABLE_JUDGES=true
+ENABLE_SUPERVISOR=true
+ENABLE_OUTPUT_SUPERVISOR=true
+ENABLE_PARALLEL_GUARDRAILS=true
+GUARDRAILS_FAIL_FAST=true
+OUTPUT_SUPERVISOR_MAX_RETRIES=3
+GUARDRAILS_CONFIG_PATH=./config/guardrails.yaml
+JUDGES_CONFIG_PATH=./config/judges.yaml
+PROMPT_POLICY_PATH=./config/prompt_policy.yaml
+
+###############################################################################
+# Gateway de canais
+###############################################################################
+DEFAULT_CHANNEL=web
+# embedded = backend may parse simple/native channel payloads.
+# external = backend only accepts GatewayRequest normalized by an external Channel Gateway.
+FRAMEWORK_CHANNEL_INPUT_MODE=embedded
+ENABLE_VOICE_ADAPTER=true
+ENABLE_WHATSAPP_ADAPTER=true
+ENABLE_TEXT_ADAPTER=true
+
+#################################################
+# ENTERPRISE ROUTING
+#################################################
+# Arquivo YAML com intents, keywords, políticas de estado e fallback.
+ROUTING_CONFIG_PATH=./config/routing.yaml
+# true = usa LLM para classificar quando keywords/estado não resolverem.
+# Em produção, costuma ser útil; em desenvolvimento, false evita custo e latência.
+ENABLE_LLM_ROUTER=true
+
+# Semantic route stickiness (optional).
+# Uses a lightweight LLM profile to decide only CONTINUE vs ROUTE.
+# There are no regexes or deterministic language rules.
+ENABLE_ROUTE_STICKINESS=false
+ROUTE_STICKINESS_LLM_PROFILE=route_continuity
+ROUTE_STICKINESS_CONFIDENCE_THRESHOLD=0.90
+ROUTE_STICKINESS_HISTORY_TURNS=2
+ROUTE_STICKINESS_MAX_TOKENS=80
+HUMAN_HANDOFF_MESSAGE=Vou encaminhar seu atendimento para uma pessoa.
+END_SESSION_MESSAGE=Atendimento encerrado. Obrigado pelo contato.
+
+###############################################################################
+# MCP / Tools
+###############################################################################
+ENABLE_MCP_TOOLS=true
+MCP_SERVERS_CONFIG_PATH=./config/mcp_servers.yaml
+TOOLS_CONFIG_PATH=./config/tools.yaml
+TOOL_POLICIES_PATH=./config/tool_policies.yaml
+MCP_TOOL_TIMEOUT_SECONDS=30
+
+# router = EnterpriseRouter seleciona um agente; supervisor = pode acionar múltiplos agentes
+ROUTING_MODE=router
+
+# Usage/cost accounting
+USAGE_REPOSITORY_PROVIDER=autonomous
+IDENTITY_CONFIG_PATH=./config/identity.yaml
+MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml
+
+# -----------------------------------------------------------------------------
+# ConversationSummaryMemory / compressão de contexto conversacional
+# -----------------------------------------------------------------------------
+ENABLE_CONVERSATION_SUMMARY_MEMORY=true
+MEMORY_CONTEXT_STRATEGY=summary
+MEMORY_HISTORY_LIMIT=80
+MEMORY_RECENT_MESSAGES_LIMIT=8
+MEMORY_SUMMARY_TRIGGER_MESSAGES=20
+MEMORY_MAX_SUMMARY_CHARS=6000
+MEMORY_SUMMARY_USE_LLM=true
+MEMORY_INJECT_RECENT_MESSAGES=true
+MEMORY_INJECT_SUMMARY=true
+
+###############################################################################
+# MCP Gateway
+###############################################################################
+# true = framework routes tool calls to the dedicated MCP Gateway.
+# false = framework calls MCP servers directly from mcp_servers.yaml.
+MCP_GATEWAY_ENABLED=true
+MCP_GATEWAY_URL=http://localhost:8300
+MCP_GATEWAY_TIMEOUT_SECONDS=60
+# MCP_GATEWAY_TOKEN=
+MCP_GATEWAY_AGENT_ID=telecom_contas
+MCP_GATEWAY_TENANT_ID=default
+
+###############################################################################
+# LONG-TERM MEMORY
+###############################################################################
+ENABLE_LONG_TERM_MEMORY=true
+LONG_TERM_MEMORY_PROVIDER=sqlite
+LONG_TERM_MEMORY_SQLITE_PATH=./data/agent_framework.db
+LONG_TERM_MEMORY_TABLE=agentfw_long_term_memory
+# For Autonomous/Oracle, defaults to ${ADB_TABLE_PREFIX}_LONG_TERM_MEMORY
+# LONG_TERM_MEMORY_ORACLE_TABLE=AGENTFW_LONG_TERM_MEMORY
+LONG_TERM_MEMORY_MAX_CONTEXT_ITEMS=20
+LONG_TERM_MEMORY_MIN_CONFIDENCE=0.70
+LONG_TERM_MEMORY_AUTO_EXTRACT=true
+LONG_TERM_MEMORY_INJECT_CONTEXT=true

Tuning-Performance/Normal/templates/agent_template_backend/.env.example

@@ -0,0 +1,195 @@
+###############################################################################
+# AI AGENT PLATFORM - CONFIGURAÇÃO ÚNICA
+# Este arquivo é lido por Pydantic Settings no framework e no backend template.
+###############################################################################
+
+APP_NAME=ai-agent-template
+APP_ENV=local
+LOG_LEVEL=INFO
+API_HOST=0.0.0.0
+API_PORT=8000
+CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173
+
+###############################################################################
+# LLM - OCI Generative AI como provider principal
+###############################################################################
+# Opções: mock, oci_openai, oci_sdk, openai_compatible
+LLM_PROVIDER=oci_openai
+LLM_TEMPERATURE=0.2
+LLM_MAX_TOKENS=2048
+LLM_TIMEOUT_SECONDS=120
+
+# OCI OpenAI-compatible endpoint
+OCI_GENAI_BASE_URL=https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1
+OCI_GENAI_MODEL=openai.gpt-4.1
+OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
+OCI_GENAI_PROJECT_OCID=
+
+# OCI SDK / signer / profiles
+OCI_CONFIG_FILE=~/.oci/config
+OCI_PROFILE=DEFAULT
+OCI_COMPARTMENT_ID=ocid1.compartment.oc1..aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
+OCI_REGION=us-chicago-1
+
+###############################################################################
+# Persistência
+###############################################################################
+# Opções: memory, autonomous, mongodb
+SESSION_REPOSITORY_PROVIDER=sqlite
+MEMORY_REPOSITORY_PROVIDER=sqlite
+CHECKPOINT_REPOSITORY_PROVIDER=sqlite
+SQLITE_DB_PATH=./data/agent_framework.db
+
+# Autonomous Database
+ADB_USER=admin
+ADB_PASSWORD=fjhsdf04954hf
+ADB_DSN=oradb23aidev_high
+ADB_WALLET_LOCATION=/ORACLE/DEFAULT/Wallet_ORADB23aiDev
+ADB_WALLET_PASSWORD=fjhsdf04954hf
+ADB_TABLE_PREFIX=AGENTFW
+
+# MongoDB - também pode representar Autonomous usando API compatível com Mongo, se habilitada no ambiente
+MONGODB_URI=mongodb://mongo:mongopassword@localhost:27017
+MONGODB_DATABASE=agent_platform
+
+# Redis
+REDIS_URL=redis://localhost:6379/0
+ENABLE_REDIS_CACHE=false
+
+###############################################################################
+# RAG / Vector / Graph
+###############################################################################
+VECTOR_STORE_PROVIDER=sqlite
+GRAPH_STORE_PROVIDER=sqlite
+RAG_TOP_K=5
+EMBEDDING_PROVIDER=mock
+OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
+RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
+
+###############################################################################
+# Observabilidade
+###############################################################################
+ENABLE_LANGFUSE=true
+LANGFUSE_TRACE_MODE=compact # Opcional: verbose, compact
+LANGFUSE_PUBLIC_KEY=pk-lf-2f9da109-5b0f-4c78-b61d-9598ed787eba
+LANGFUSE_SECRET_KEY=sk-lf-a4cb0cdd-f2ea-4468-9911-cebeb91ba944
+LANGFUSE_HOST=http://localhost:3005
+ENABLE_OTEL=false
+OTEL_EXPORTER_OTLP_ENDPOINT=
+OTEL_SERVICE_NAME=ai-agent-template
+ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
+
+###############################################################################
+# Analytics / Observer corporativo
+###############################################################################
+# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
+ENABLE_ANALYTICS=false
+# Providers aceitos: oci_streaming,pubsub,noop
+ANALYTICS_PROVIDERS=pubsub
+# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
+AGENT_PUBSUB_TOPIC=
+GCP_PUBSUB_TOPIC_PATH=
+GCP_PROJECT_ID=
+GCP_PUBSUB_TOPIC=
+GCP_PUBSUB_TIMEOUT_SECONDS=30
+# Credencial GCP segue padrão Google:
+# GOOGLE_APPLICATION_CREDENTIALS=/secrets/gcp-service-account.json
+
+###############################################################################
+# OCI Streaming
+###############################################################################
+ENABLE_OCI_STREAMING=false
+OCI_STREAM_ENDPOINT=
+OCI_STREAM_OCID=
+OCI_STREAM_PARTITION_KEY=agent-events
+
+###############################################################################
+# Guardrails, Judges, Supervisor
+###############################################################################
+ENABLE_INPUT_GUARDRAILS=true
+ENABLE_OUTPUT_GUARDRAILS=true
+ENABLE_JUDGES=true
+ENABLE_SUPERVISOR=true
+ENABLE_OUTPUT_SUPERVISOR=true
+ENABLE_PARALLEL_GUARDRAILS=true
+GUARDRAILS_FAIL_FAST=true
+OUTPUT_SUPERVISOR_MAX_RETRIES=3
+GUARDRAILS_CONFIG_PATH=./config/guardrails.yaml
+JUDGES_CONFIG_PATH=./config/judges.yaml
+PROMPT_POLICY_PATH=./config/prompt_policy.yaml
+
+###############################################################################
+# Gateway de canais
+###############################################################################
+DEFAULT_CHANNEL=web
+# embedded = backend may parse simple/native channel payloads.
+# external = backend only accepts GatewayRequest normalized by an external Channel Gateway.
+FRAMEWORK_CHANNEL_INPUT_MODE=embedded
+ENABLE_VOICE_ADAPTER=true
+ENABLE_WHATSAPP_ADAPTER=true
+ENABLE_TEXT_ADAPTER=true
+
+#################################################
+# ENTERPRISE ROUTING
+#################################################
+# Arquivo YAML com intents, keywords, políticas de estado e fallback.
+ROUTING_CONFIG_PATH=./config/routing.yaml
+# true = usa LLM para classificar quando keywords/estado não resolverem.
+# Em produção, costuma ser útil; em desenvolvimento, false evita custo e latência.
+ENABLE_LLM_ROUTER=true
+
+###############################################################################
+# MCP / Tools
+###############################################################################
+ENABLE_MCP_TOOLS=true
+MCP_SERVERS_CONFIG_PATH=./config/mcp_servers.yaml
+TOOLS_CONFIG_PATH=./config/tools.yaml
+TOOL_POLICIES_PATH=./config/tool_policies.yaml
+MCP_TOOL_TIMEOUT_SECONDS=30
+
+# router = EnterpriseRouter seleciona um agente; supervisor = pode acionar múltiplos agentes
+ROUTING_MODE=router
+
+# Usage/cost accounting
+USAGE_REPOSITORY_PROVIDER=sqlite
+IDENTITY_CONFIG_PATH=./config/identity.yaml
+MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml
+
+# -----------------------------------------------------------------------------
+# ConversationSummaryMemory / compressão de contexto conversacional
+# -----------------------------------------------------------------------------
+ENABLE_CONVERSATION_SUMMARY_MEMORY=true
+MEMORY_CONTEXT_STRATEGY=summary
+MEMORY_HISTORY_LIMIT=80
+MEMORY_RECENT_MESSAGES_LIMIT=8
+MEMORY_SUMMARY_TRIGGER_MESSAGES=20
+MEMORY_MAX_SUMMARY_CHARS=6000
+MEMORY_SUMMARY_USE_LLM=true
+MEMORY_INJECT_RECENT_MESSAGES=true
+MEMORY_INJECT_SUMMARY=true
+
+###############################################################################
+# MCP Gateway
+###############################################################################
+# true = framework routes tool calls to the dedicated MCP Gateway.
+# false = framework calls MCP servers directly from mcp_servers.yaml.
+MCP_GATEWAY_ENABLED=true
+MCP_GATEWAY_URL=http://localhost:8300
+MCP_GATEWAY_TIMEOUT_SECONDS=60
+# MCP_GATEWAY_TOKEN=
+MCP_GATEWAY_AGENT_ID=telecom_contas
+MCP_GATEWAY_TENANT_ID=default
+
+###############################################################################
+# LONG-TERM MEMORY
+###############################################################################
+ENABLE_LONG_TERM_MEMORY=true
+LONG_TERM_MEMORY_PROVIDER=sqlite
+LONG_TERM_MEMORY_SQLITE_PATH=./data/agent_framework.db
+LONG_TERM_MEMORY_TABLE=agentfw_long_term_memory
+# For Autonomous/Oracle, defaults to ${ADB_TABLE_PREFIX}_LONG_TERM_MEMORY
+# LONG_TERM_MEMORY_ORACLE_TABLE=AGENTFW_LONG_TERM_MEMORY
+LONG_TERM_MEMORY_MAX_CONTEXT_ITEMS=20
+LONG_TERM_MEMORY_MIN_CONFIDENCE=0.70
+LONG_TERM_MEMORY_AUTO_EXTRACT=true
+LONG_TERM_MEMORY_INJECT_CONTEXT=true

Tuning-Performance/Normal/templates/agent_template_backend_day_zero/.env.example

@@ -0,0 +1,192 @@
+###############################################################################
+# AI AGENT PLATFORM - CONFIGURAÇÃO ÚNICA
+# Este arquivo é lido por Pydantic Settings no framework e no backend template.
+###############################################################################
+
+APP_NAME=ai-agent-template
+APP_ENV=local
+LOG_LEVEL=INFO
+API_HOST=0.0.0.0
+API_PORT=8000
+CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173
+
+###############################################################################
+# LLM - OCI Generative AI como provider principal
+###############################################################################
+# Opções: mock, oci_openai, oci_sdk, openai_compatible
+LLM_PROVIDER=oci_openai
+LLM_TEMPERATURE=0.2
+LLM_MAX_TOKENS=2048
+LLM_TIMEOUT_SECONDS=120
+
+# OCI OpenAI-compatible endpoint
+OCI_GENAI_BASE_URL=https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1
+OCI_GENAI_MODEL=openai.gpt-4.1
+OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
+OCI_GENAI_PROJECT_OCID=
+
+# OCI SDK / signer / profiles
+OCI_CONFIG_FILE=~/.oci/config
+OCI_PROFILE=DEFAULT
+OCI_COMPARTMENT_ID=ocid1.compartment.oc1..aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
+OCI_REGION=us-chicago-1
+
+###############################################################################
+# Persistência
+###############################################################################
+# Opções: memory, autonomous, mongodb
+SESSION_REPOSITORY_PROVIDER=sqlite
+MEMORY_REPOSITORY_PROVIDER=sqlite
+CHECKPOINT_REPOSITORY_PROVIDER=sqlite
+SQLITE_DB_PATH=./data/agent_framework.db
+
+# Autonomous Database
+ADB_USER=admin
+ADB_PASSWORD=fjhsdf04954hf
+ADB_DSN=oradb23aidev_high
+ADB_WALLET_LOCATION=/ORACLE/DEFAULT/Wallet_ORADB23aiDev
+ADB_WALLET_PASSWORD=fjhsdf04954hf
+ADB_TABLE_PREFIX=AGENTFW
+
+# MongoDB - também pode representar Autonomous usando API compatível com Mongo, se habilitada no ambiente
+MONGODB_URI=mongodb://mongo:mongopassword@localhost:27017
+MONGODB_DATABASE=agent_platform
+
+# Redis
+REDIS_URL=redis://localhost:6379/0
+ENABLE_REDIS_CACHE=false
+
+###############################################################################
+# RAG / Vector / Graph
+###############################################################################
+VECTOR_STORE_PROVIDER=sqlite
+GRAPH_STORE_PROVIDER=sqlite
+RAG_TOP_K=5
+EMBEDDING_PROVIDER=mock
+OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
+RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
+
+###############################################################################
+# Observabilidade
+###############################################################################
+ENABLE_LANGFUSE=true
+LANGFUSE_TRACE_MODE=compact # Opcional: verbose, compact
+LANGFUSE_PUBLIC_KEY=pk-lf-2f9da109-5b0f-4c78-b61d-9598ed787eba
+LANGFUSE_SECRET_KEY=sk-lf-a4cb0cdd-f2ea-4468-9911-cebeb91ba944
+LANGFUSE_HOST=http://localhost:3005
+ENABLE_OTEL=false
+OTEL_EXPORTER_OTLP_ENDPOINT=
+OTEL_SERVICE_NAME=ai-agent-template
+ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
+
+###############################################################################
+# Analytics / Observer corporativo
+###############################################################################
+# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
+ENABLE_ANALYTICS=false
+# Providers aceitos: oci_streaming,pubsub,noop
+ANALYTICS_PROVIDERS=pubsub
+# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
+AGENT_PUBSUB_TOPIC=
+GCP_PUBSUB_TOPIC_PATH=
+GCP_PROJECT_ID=
+GCP_PUBSUB_TOPIC=
+GCP_PUBSUB_TIMEOUT_SECONDS=30
+# Credencial GCP segue padrão Google:
+# GOOGLE_APPLICATION_CREDENTIALS=/secrets/gcp-service-account.json
+
+###############################################################################
+# OCI Streaming
+###############################################################################
+ENABLE_OCI_STREAMING=false
+OCI_STREAM_ENDPOINT=
+OCI_STREAM_OCID=
+OCI_STREAM_PARTITION_KEY=agent-events
+
+###############################################################################
+# Guardrails, Judges, Supervisor
+###############################################################################
+ENABLE_INPUT_GUARDRAILS=true
+ENABLE_OUTPUT_GUARDRAILS=true
+ENABLE_JUDGES=true
+ENABLE_SUPERVISOR=true
+ENABLE_OUTPUT_SUPERVISOR=true
+ENABLE_PARALLEL_GUARDRAILS=true
+GUARDRAILS_FAIL_FAST=true
+OUTPUT_SUPERVISOR_MAX_RETRIES=3
+GUARDRAILS_CONFIG_PATH=./config/guardrails.yaml
+JUDGES_CONFIG_PATH=./config/judges.yaml
+PROMPT_POLICY_PATH=./config/prompt_policy.yaml
+
+###############################################################################
+# Gateway de canais
+###############################################################################
+DEFAULT_CHANNEL=web
+ENABLE_VOICE_ADAPTER=true
+ENABLE_WHATSAPP_ADAPTER=true
+ENABLE_TEXT_ADAPTER=true
+
+#################################################
+# ENTERPRISE ROUTING
+#################################################
+# Arquivo YAML com intents, keywords, políticas de estado e fallback.
+ROUTING_CONFIG_PATH=./config/routing.yaml
+# true = usa LLM para classificar quando keywords/estado não resolverem.
+# Em produção, costuma ser útil; em desenvolvimento, false evita custo e latência.
+ENABLE_LLM_ROUTER=true
+
+###############################################################################
+# MCP / Tools
+###############################################################################
+ENABLE_MCP_TOOLS=true
+MCP_SERVERS_CONFIG_PATH=./config/mcp_servers.yaml
+TOOLS_CONFIG_PATH=./config/tools.yaml
+TOOL_POLICIES_PATH=./config/tool_policies.yaml
+MCP_TOOL_TIMEOUT_SECONDS=30
+
+# router = EnterpriseRouter seleciona um agente; supervisor = pode acionar múltiplos agentes
+ROUTING_MODE=router
+
+# Usage/cost accounting
+USAGE_REPOSITORY_PROVIDER=sqlite
+IDENTITY_CONFIG_PATH=./config/identity.yaml
+MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml
+
+# -----------------------------------------------------------------------------
+# ConversationSummaryMemory / compressão de contexto conversacional
+# -----------------------------------------------------------------------------
+ENABLE_CONVERSATION_SUMMARY_MEMORY=true
+MEMORY_CONTEXT_STRATEGY=summary
+MEMORY_HISTORY_LIMIT=80
+MEMORY_RECENT_MESSAGES_LIMIT=8
+MEMORY_SUMMARY_TRIGGER_MESSAGES=20
+MEMORY_MAX_SUMMARY_CHARS=6000
+MEMORY_SUMMARY_USE_LLM=true
+MEMORY_INJECT_RECENT_MESSAGES=true
+MEMORY_INJECT_SUMMARY=true
+
+###############################################################################
+# MCP Gateway
+###############################################################################
+# true = framework routes tool calls to the dedicated MCP Gateway.
+# false = framework calls MCP servers directly from mcp_servers.yaml.
+MCP_GATEWAY_ENABLED=true
+MCP_GATEWAY_URL=http://localhost:8300
+MCP_GATEWAY_TIMEOUT_SECONDS=60
+# MCP_GATEWAY_TOKEN=
+MCP_GATEWAY_AGENT_ID=telecom_contas
+MCP_GATEWAY_TENANT_ID=default
+
+###############################################################################
+# LONG-TERM MEMORY
+###############################################################################
+ENABLE_LONG_TERM_MEMORY=true
+LONG_TERM_MEMORY_PROVIDER=sqlite
+LONG_TERM_MEMORY_SQLITE_PATH=./data/agent_framework.db
+LONG_TERM_MEMORY_TABLE=agentfw_long_term_memory
+# For Autonomous/Oracle, defaults to ${ADB_TABLE_PREFIX}_LONG_TERM_MEMORY
+# LONG_TERM_MEMORY_ORACLE_TABLE=AGENTFW_LONG_TERM_MEMORY
+LONG_TERM_MEMORY_MAX_CONTEXT_ITEMS=20
+LONG_TERM_MEMORY_MIN_CONFIDENCE=0.70
+LONG_TERM_MEMORY_AUTO_EXTRACT=true
+LONG_TERM_MEMORY_INJECT_CONTEXT=true

Tuning-Performance/Route_Stickness/templates/agent_template_backend/.env.example

@@ -0,0 +1,207 @@
+###############################################################################
+# AI AGENT PLATFORM - CONFIGURAÇÃO ÚNICA
+# Este arquivo é lido por Pydantic Settings no framework e no backend template.
+###############################################################################
+
+APP_NAME=ai-agent-template
+APP_ENV=local
+LOG_LEVEL=INFO
+API_HOST=0.0.0.0
+API_PORT=8000
+CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173
+
+###############################################################################
+# LLM - OCI Generative AI como provider principal
+###############################################################################
+# Opções: mock, oci_openai, oci_sdk, openai_compatible
+LLM_PROVIDER=oci_openai
+LLM_TEMPERATURE=0.2
+LLM_MAX_TOKENS=2048
+LLM_TIMEOUT_SECONDS=120
+
+# OCI OpenAI-compatible endpoint
+OCI_GENAI_BASE_URL=https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1
+OCI_GENAI_MODEL=openai.gpt-4.1
+OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
+OCI_GENAI_PROJECT_OCID=
+
+# OCI SDK / signer / profiles
+OCI_CONFIG_FILE=~/.oci/config
+OCI_PROFILE=DEFAULT
+OCI_COMPARTMENT_ID=ocid1.compartment.oc1..aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
+OCI_REGION=us-chicago-1
+
+###############################################################################
+# Persistência
+###############################################################################
+# Opções: memory, autonomous, mongodb
+SESSION_REPOSITORY_PROVIDER=sqlite
+MEMORY_REPOSITORY_PROVIDER=sqlite
+CHECKPOINT_REPOSITORY_PROVIDER=sqlite
+SQLITE_DB_PATH=./data/agent_framework.db
+
+# Autonomous Database
+ADB_USER=admin
+ADB_PASSWORD=fjhsdf04954hf
+ADB_DSN=oradb23aidev_high
+ADB_WALLET_LOCATION=/ORACLE/DEFAULT/Wallet_ORADB23aiDev
+ADB_WALLET_PASSWORD=fjhsdf04954hf
+ADB_TABLE_PREFIX=AGENTFW
+
+# MongoDB - também pode representar Autonomous usando API compatível com Mongo, se habilitada no ambiente
+MONGODB_URI=mongodb://mongo:mongopassword@localhost:27017
+MONGODB_DATABASE=agent_platform
+
+# Redis
+REDIS_URL=redis://localhost:6379/0
+ENABLE_REDIS_CACHE=false
+
+###############################################################################
+# RAG / Vector / Graph
+###############################################################################
+VECTOR_STORE_PROVIDER=sqlite
+GRAPH_STORE_PROVIDER=sqlite
+RAG_TOP_K=5
+EMBEDDING_PROVIDER=mock
+OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
+RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
+
+###############################################################################
+# Observabilidade
+###############################################################################
+ENABLE_LANGFUSE=true
+LANGFUSE_TRACE_MODE=compact # Opcional: verbose, compact
+LANGFUSE_PUBLIC_KEY=pk-lf-2f9da109-5b0f-4c78-b61d-9598ed787eba
+LANGFUSE_SECRET_KEY=sk-lf-a4cb0cdd-f2ea-4468-9911-cebeb91ba944
+LANGFUSE_HOST=http://localhost:3005
+ENABLE_OTEL=false
+OTEL_EXPORTER_OTLP_ENDPOINT=
+OTEL_SERVICE_NAME=ai-agent-template
+ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
+
+###############################################################################
+# Analytics / Observer corporativo
+###############################################################################
+# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
+ENABLE_ANALYTICS=false
+# Providers aceitos: oci_streaming,pubsub,noop
+ANALYTICS_PROVIDERS=pubsub
+# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
+AGENT_PUBSUB_TOPIC=
+GCP_PUBSUB_TOPIC_PATH=
+GCP_PROJECT_ID=
+GCP_PUBSUB_TOPIC=
+GCP_PUBSUB_TIMEOUT_SECONDS=30
+# Credencial GCP segue padrão Google:
+# GOOGLE_APPLICATION_CREDENTIALS=/secrets/gcp-service-account.json
+
+###############################################################################
+# OCI Streaming
+###############################################################################
+ENABLE_OCI_STREAMING=false
+OCI_STREAM_ENDPOINT=
+OCI_STREAM_OCID=
+OCI_STREAM_PARTITION_KEY=agent-events
+
+###############################################################################
+# Guardrails, Judges, Supervisor
+###############################################################################
+ENABLE_INPUT_GUARDRAILS=true
+ENABLE_OUTPUT_GUARDRAILS=true
+ENABLE_JUDGES=true
+ENABLE_SUPERVISOR=true
+ENABLE_OUTPUT_SUPERVISOR=true
+ENABLE_PARALLEL_GUARDRAILS=true
+GUARDRAILS_FAIL_FAST=true
+OUTPUT_SUPERVISOR_MAX_RETRIES=3
+GUARDRAILS_CONFIG_PATH=./config/guardrails.yaml
+JUDGES_CONFIG_PATH=./config/judges.yaml
+PROMPT_POLICY_PATH=./config/prompt_policy.yaml
+
+###############################################################################
+# Gateway de canais
+###############################################################################
+DEFAULT_CHANNEL=web
+# embedded = backend may parse simple/native channel payloads.
+# external = backend only accepts GatewayRequest normalized by an external Channel Gateway.
+FRAMEWORK_CHANNEL_INPUT_MODE=embedded
+ENABLE_VOICE_ADAPTER=true
+ENABLE_WHATSAPP_ADAPTER=true
+ENABLE_TEXT_ADAPTER=true
+
+#################################################
+# ENTERPRISE ROUTING
+#################################################
+# Arquivo YAML com intents, keywords, políticas de estado e fallback.
+ROUTING_CONFIG_PATH=./config/routing.yaml
+# true = usa LLM para classificar quando keywords/estado não resolverem.
+# Em produção, costuma ser útil; em desenvolvimento, false evita custo e latência.
+ENABLE_LLM_ROUTER=true
+
+# Semantic route stickiness (optional).
+# Uses a lightweight LLM profile to decide only CONTINUE vs ROUTE.
+# There are no regexes or deterministic language rules.
+ENABLE_ROUTE_STICKINESS=true
+ROUTE_STICKINESS_LLM_PROFILE=route_continuity
+ROUTE_STICKINESS_CONFIDENCE_THRESHOLD=0.90
+ROUTE_STICKINESS_HISTORY_TURNS=2
+ROUTE_STICKINESS_MAX_TOKENS=80
+HUMAN_HANDOFF_MESSAGE=Vou encaminhar seu atendimento para uma pessoa.
+END_SESSION_MESSAGE=Atendimento encerrado. Obrigado pelo contato.
+SESSION_ALREADY_ENDED_MESSAGE=Este atendimento já foi encerrado. Inicie uma nova sessão para continuar.
+
+###############################################################################
+# MCP / Tools
+###############################################################################
+ENABLE_MCP_TOOLS=true
+MCP_SERVERS_CONFIG_PATH=./config/mcp_servers.yaml
+TOOLS_CONFIG_PATH=./config/tools.yaml
+TOOL_POLICIES_PATH=./config/tool_policies.yaml
+MCP_TOOL_TIMEOUT_SECONDS=30
+
+# router = EnterpriseRouter seleciona um agente; supervisor = pode acionar múltiplos agentes
+ROUTING_MODE=router
+
+# Usage/cost accounting
+USAGE_REPOSITORY_PROVIDER=sqlite
+IDENTITY_CONFIG_PATH=./config/identity.yaml
+MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml
+
+# -----------------------------------------------------------------------------
+# ConversationSummaryMemory / compressão de contexto conversacional
+# -----------------------------------------------------------------------------
+ENABLE_CONVERSATION_SUMMARY_MEMORY=true
+MEMORY_CONTEXT_STRATEGY=summary
+MEMORY_HISTORY_LIMIT=80
+MEMORY_RECENT_MESSAGES_LIMIT=8
+MEMORY_SUMMARY_TRIGGER_MESSAGES=20
+MEMORY_MAX_SUMMARY_CHARS=6000
+MEMORY_SUMMARY_USE_LLM=true
+MEMORY_INJECT_RECENT_MESSAGES=true
+MEMORY_INJECT_SUMMARY=true
+
+###############################################################################
+# MCP Gateway
+###############################################################################
+# true = framework routes tool calls to the dedicated MCP Gateway.
+# false = framework calls MCP servers directly from mcp_servers.yaml.
+MCP_GATEWAY_ENABLED=true
+MCP_GATEWAY_URL=http://localhost:8300
+MCP_GATEWAY_TIMEOUT_SECONDS=60
+# MCP_GATEWAY_TOKEN=
+MCP_GATEWAY_AGENT_ID=telecom_contas
+MCP_GATEWAY_TENANT_ID=default
+
+###############################################################################
+# LONG-TERM MEMORY
+###############################################################################
+ENABLE_LONG_TERM_MEMORY=true
+LONG_TERM_MEMORY_PROVIDER=sqlite
+LONG_TERM_MEMORY_SQLITE_PATH=./data/agent_framework.db
+LONG_TERM_MEMORY_TABLE=agentfw_long_term_memory
+# For Autonomous/Oracle, defaults to ${ADB_TABLE_PREFIX}_LONG_TERM_MEMORY
+# LONG_TERM_MEMORY_ORACLE_TABLE=AGENTFW_LONG_TERM_MEMORY
+LONG_TERM_MEMORY_MAX_CONTEXT_ITEMS=20
+LONG_TERM_MEMORY_MIN_CONFIDENCE=0.70
+LONG_TERM_MEMORY_AUTO_EXTRACT=true
+LONG_TERM_MEMORY_INJECT_CONTEXT=true

Tuning-Performance/Route_Stickness/templates/agent_template_backend_day_zero/.env.example

@@ -0,0 +1,202 @@
+###############################################################################
+# AI AGENT PLATFORM - CONFIGURAÇÃO ÚNICA
+# Este arquivo é lido por Pydantic Settings no framework e no backend template.
+###############################################################################
+
+APP_NAME=ai-agent-template
+APP_ENV=local
+LOG_LEVEL=INFO
+API_HOST=0.0.0.0
+API_PORT=8000
+CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173
+
+###############################################################################
+# LLM - OCI Generative AI como provider principal
+###############################################################################
+# Opções: mock, oci_openai, oci_sdk, openai_compatible
+LLM_PROVIDER=oci_openai
+LLM_TEMPERATURE=0.2
+LLM_MAX_TOKENS=2048
+LLM_TIMEOUT_SECONDS=120
+
+# OCI OpenAI-compatible endpoint
+OCI_GENAI_BASE_URL=https://inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1
+OCI_GENAI_MODEL=openai.gpt-4.1
+OCI_GENAI_API_KEY=sk-ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6ph3FgX6
+OCI_GENAI_PROJECT_OCID=
+
+# OCI SDK / signer / profiles
+OCI_CONFIG_FILE=~/.oci/config
+OCI_PROFILE=DEFAULT
+OCI_COMPARTMENT_ID=ocid1.compartment.oc1..aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
+OCI_REGION=us-chicago-1
+
+###############################################################################
+# Persistência
+###############################################################################
+# Opções: memory, autonomous, mongodb
+SESSION_REPOSITORY_PROVIDER=sqlite
+MEMORY_REPOSITORY_PROVIDER=sqlite
+CHECKPOINT_REPOSITORY_PROVIDER=sqlite
+SQLITE_DB_PATH=./data/agent_framework.db
+
+# Autonomous Database
+ADB_USER=admin
+ADB_PASSWORD=fjhsdf04954hf
+ADB_DSN=oradb23aidev_high
+ADB_WALLET_LOCATION=/ORACLE/DEFAULT/Wallet_ORADB23aiDev
+ADB_WALLET_PASSWORD=fjhsdf04954hf
+ADB_TABLE_PREFIX=AGENTFW
+
+# MongoDB - também pode representar Autonomous usando API compatível com Mongo, se habilitada no ambiente
+MONGODB_URI=mongodb://mongo:mongopassword@localhost:27017
+MONGODB_DATABASE=agent_platform
+
+# Redis
+REDIS_URL=redis://localhost:6379/0
+ENABLE_REDIS_CACHE=false
+
+###############################################################################
+# RAG / Vector / Graph
+###############################################################################
+VECTOR_STORE_PROVIDER=sqlite
+GRAPH_STORE_PROVIDER=sqlite
+RAG_TOP_K=5
+EMBEDDING_PROVIDER=mock
+OCI_EMBEDDING_MODEL=cohere.embed-multilingual-v3.0
+RAG_FILE_GLOBS=*.md,*.txt,*.yaml,*.yml,*.json
+
+###############################################################################
+# Observabilidade
+###############################################################################
+ENABLE_LANGFUSE=true
+LANGFUSE_TRACE_MODE=compact # Opcional: verbose, compact
+LANGFUSE_PUBLIC_KEY=pk-lf-2f9da109-5b0f-4c78-b61d-9598ed787eba
+LANGFUSE_SECRET_KEY=sk-lf-a4cb0cdd-f2ea-4468-9911-cebeb91ba944
+LANGFUSE_HOST=http://localhost:3005
+ENABLE_OTEL=false
+OTEL_EXPORTER_OTLP_ENDPOINT=
+OTEL_SERVICE_NAME=ai-agent-template
+ENABLE_LANGFUSE_OPENAI_AUTO_INSTRUMENTATION=true
+
+###############################################################################
+# Analytics / Observer corporativo
+###############################################################################
+# Quando true, AgentObserver publica eventos IC.*, NOC.* e GRL.* nos providers abaixo.
+ENABLE_ANALYTICS=false
+# Providers aceitos: oci_streaming,pubsub,noop
+ANALYTICS_PROVIDERS=pubsub
+# Compatibilidade FIRST/TIM: pode informar AGENT_PUBSUB_TOPIC diretamente.
+AGENT_PUBSUB_TOPIC=
+GCP_PUBSUB_TOPIC_PATH=
+GCP_PROJECT_ID=
+GCP_PUBSUB_TOPIC=
+GCP_PUBSUB_TIMEOUT_SECONDS=30
+# Credencial GCP segue padrão Google:
+# GOOGLE_APPLICATION_CREDENTIALS=/secrets/gcp-service-account.json
+
+###############################################################################
+# OCI Streaming
+###############################################################################
+ENABLE_OCI_STREAMING=false
+OCI_STREAM_ENDPOINT=
+OCI_STREAM_OCID=
+OCI_STREAM_PARTITION_KEY=agent-events
+
+###############################################################################
+# Guardrails, Judges, Supervisor
+###############################################################################
+ENABLE_INPUT_GUARDRAILS=true
+ENABLE_OUTPUT_GUARDRAILS=true
+ENABLE_JUDGES=true
+ENABLE_SUPERVISOR=true
+ENABLE_OUTPUT_SUPERVISOR=true
+ENABLE_PARALLEL_GUARDRAILS=true
+GUARDRAILS_FAIL_FAST=true
+OUTPUT_SUPERVISOR_MAX_RETRIES=3
+GUARDRAILS_CONFIG_PATH=./config/guardrails.yaml
+JUDGES_CONFIG_PATH=./config/judges.yaml
+PROMPT_POLICY_PATH=./config/prompt_policy.yaml
+
+###############################################################################
+# Gateway de canais
+###############################################################################
+DEFAULT_CHANNEL=web
+ENABLE_VOICE_ADAPTER=true
+ENABLE_WHATSAPP_ADAPTER=true
+ENABLE_TEXT_ADAPTER=true
+
+#################################################
+# ENTERPRISE ROUTING
+#################################################
+# Arquivo YAML com intents, keywords, políticas de estado e fallback.
+ROUTING_CONFIG_PATH=./config/routing.yaml
+# true = usa LLM para classificar quando keywords/estado não resolverem.
+# Em produção, costuma ser útil; em desenvolvimento, false evita custo e latência.
+ENABLE_LLM_ROUTER=true
+
+# Continuidade semântica, handoff humano e encerramento global.
+ENABLE_ROUTE_STICKINESS=true
+ROUTE_STICKINESS_LLM_PROFILE=route_continuity
+ROUTE_STICKINESS_CONFIDENCE_THRESHOLD=0.90
+ROUTE_STICKINESS_HISTORY_TURNS=2
+ROUTE_STICKINESS_MAX_TOKENS=80
+HUMAN_HANDOFF_MESSAGE=Vou encaminhar seu atendimento para uma pessoa.
+END_SESSION_MESSAGE=Atendimento encerrado. Obrigado pelo contato.
+SESSION_ALREADY_ENDED_MESSAGE=Este atendimento já foi encerrado. Inicie uma nova sessão para continuar.
+
+###############################################################################
+# MCP / Tools
+###############################################################################
+ENABLE_MCP_TOOLS=true
+MCP_SERVERS_CONFIG_PATH=./config/mcp_servers.yaml
+TOOLS_CONFIG_PATH=./config/tools.yaml
+TOOL_POLICIES_PATH=./config/tool_policies.yaml
+MCP_TOOL_TIMEOUT_SECONDS=30
+
+# router = EnterpriseRouter seleciona um agente; supervisor = pode acionar múltiplos agentes
+ROUTING_MODE=router
+
+# Usage/cost accounting
+USAGE_REPOSITORY_PROVIDER=sqlite
+IDENTITY_CONFIG_PATH=./config/identity.yaml
+MCP_PARAMETER_MAPPING_PATH=./config/mcp_parameter_mapping.yaml
+
+# -----------------------------------------------------------------------------
+# ConversationSummaryMemory / compressão de contexto conversacional
+# -----------------------------------------------------------------------------
+ENABLE_CONVERSATION_SUMMARY_MEMORY=true
+MEMORY_CONTEXT_STRATEGY=summary
+MEMORY_HISTORY_LIMIT=80
+MEMORY_RECENT_MESSAGES_LIMIT=8
+MEMORY_SUMMARY_TRIGGER_MESSAGES=20
+MEMORY_MAX_SUMMARY_CHARS=6000
+MEMORY_SUMMARY_USE_LLM=true
+MEMORY_INJECT_RECENT_MESSAGES=true
+MEMORY_INJECT_SUMMARY=true
+
+###############################################################################
+# MCP Gateway
+###############################################################################
+# true = framework routes tool calls to the dedicated MCP Gateway.
+# false = framework calls MCP servers directly from mcp_servers.yaml.
+MCP_GATEWAY_ENABLED=true
+MCP_GATEWAY_URL=http://localhost:8300
+MCP_GATEWAY_TIMEOUT_SECONDS=60
+# MCP_GATEWAY_TOKEN=
+MCP_GATEWAY_AGENT_ID=telecom_contas
+MCP_GATEWAY_TENANT_ID=default
+
+###############################################################################
+# LONG-TERM MEMORY
+###############################################################################
+ENABLE_LONG_TERM_MEMORY=true
+LONG_TERM_MEMORY_PROVIDER=sqlite
+LONG_TERM_MEMORY_SQLITE_PATH=./data/agent_framework.db
+LONG_TERM_MEMORY_TABLE=agentfw_long_term_memory
+# For Autonomous/Oracle, defaults to ${ADB_TABLE_PREFIX}_LONG_TERM_MEMORY
+# LONG_TERM_MEMORY_ORACLE_TABLE=AGENTFW_LONG_TERM_MEMORY
+LONG_TERM_MEMORY_MAX_CONTEXT_ITEMS=20
+LONG_TERM_MEMORY_MIN_CONFIDENCE=0.70
+LONG_TERM_MEMORY_AUTO_EXTRACT=true
+LONG_TERM_MEMORY_INJECT_CONTEXT=true

docs/LLM_RICH_RESPONSE.md

@@ -0,0 +1,74 @@
+# LLM Rich Response (`ainvoke_response`)
+
+## Objetivo
+
+O framework mantém `ainvoke()` como API retrocompatível, retornando apenas `str`, e adiciona `ainvoke_response()` para consumidores que precisam de metadados adicionais da inferência, incluindo `reasoning_content` quando o modelo/provider/API o disponibilizar.
+
+## APIs
+
+### API legada — sem alteração
+
+```python
+answer = await llm.ainvoke(messages)
+assert isinstance(answer, str)
+```
+
+Nenhum agente existente precisa ser alterado.
+
+### Nova API rica — opt-in
+
+```python
+response = await llm.ainvoke_response(messages)
+
+answer = response.content
+reasoning = response.reasoning_content
+usage = response.usage
+model = response.model
+provider = response.provider
+```
+
+`reasoning_content` é `str | None`. `None` é o comportamento esperado quando o modelo, provider ou API não expõe reasoning textual.
+
+## Backoffice
+
+Um consumidor que antes fazia:
+
+```python
+answer = await llm.ainvoke(messages)
+template = extract_response(answer)
+```
+
+pode passar a fazer:
+
+```python
+response = await llm.ainvoke_response(messages)
+template = extract_response(response.content)
+reasoning_content = response.reasoning_content
+```
+
+A lógica que espera texto continua recebendo `response.content`; o reasoning fica separado e não contamina resposta, cache, memória, judges ou guardrails.
+
+## Compatibilidade de providers customizados
+
+`LLMProvider.ainvoke_response()` possui fallback. Um provider externo que implemente apenas `ainvoke()` continua funcionando e recebe automaticamente um `LLMResponse(content=<texto>)`, com `reasoning_content=None`.
+
+Providers nativos (`mock`, OpenAI-compatible/OCI OpenAI e OCI SDK) implementam a resposta rica e tentam preservar reasoning quando presente.
+
+## Garantias de compatibilidade
+
+- `ainvoke()` continua retornando `str`.
+- Nenhum router, judge, RAG, memória, cache ou runtime existente foi migrado para a nova API.
+- `reasoning_content` nunca é fabricado pelo framework.
+- Ausência de reasoning não gera erro.
+- O output existente de telemetria continua sendo o conteúdo final, sem anexar reasoning automaticamente.
+
+## Testes
+
+Os testes específicos estão em `tests/unit/test_llm_rich_response.py` e verificam:
+
+1. provider legado que só implementa `ainvoke()`;
+2. manutenção do retorno `str` em `ainvoke()`;
+3. retorno de `LLMResponse` em `ainvoke_response()`;
+4. reasoning via atributo direto;
+5. reasoning via `model_extra`;
+6. ausência de reasoning e extração no formato OCI SDK.

libs/agent_framework/src/agent_framework/llm/__init__.py

@@ -0,0 +1,4 @@
+from .base import LLMProvider
+from .types import LLMResponse
+
+__all__ = ["LLMProvider", "LLMResponse"]

libs/agent_framework/src/agent_framework/llm/base.py

@@ -1,6 +1,25 @@
 from abc import ABC, abstractmethod
 from typing import Any
 
+from .types import LLMResponse
+
+
 class LLMProvider(ABC):
     @abstractmethod
-    async def ainvoke(self, messages: list[dict[str, str]], **kwargs: Any) -> str: ...
+    async def ainvoke(self, messages: list[dict[str, str]], **kwargs: Any) -> str:
+        """Legacy API. Must keep returning only the textual answer."""
+        ...
+
+    async def ainvoke_response(
+        self,
+        messages: list[dict[str, str]],
+        **kwargs: Any,
+    ) -> LLMResponse:
+        """Rich opt-in API with a backward-compatible fallback.
+
+        Custom providers that only implement ``ainvoke`` continue to work. They
+        simply expose ``content`` and leave optional provider metadata/reasoning
+        empty until they choose to override this method.
+        """
+        content = await self.ainvoke(messages, **kwargs)
+        return LLMResponse(content=str(content or ""))

libs/agent_framework/src/agent_framework/llm/providers.py

@@ -5,6 +5,7 @@ import os
 from typing import Any
 
 from .base import LLMProvider
+from .types import LLMResponse
 from .profile_resolver import LLMProfileResolver
 from agent_framework.observability.token_cost import TokenUsageCollector
 from agent_framework.billing.usage_repository import UsageRepository, UsageRecord
@@ -12,6 +13,62 @@ from agent_framework.billing.usage_repository import UsageRepository, UsageRecor
 logger = logging.getLogger("agent_framework.llm")
 
 
+def _coerce_reasoning_text(value: Any) -> str | None:
+    """Normalize provider-specific reasoning payloads without inventing content."""
+    if value is None:
+        return None
+    if isinstance(value, str):
+        value = value.strip()
+        return value or None
+    if isinstance(value, (list, tuple)):
+        chunks: list[str] = []
+        for item in value:
+            if isinstance(item, str):
+                text = item
+            else:
+                text = getattr(item, "text", None) or getattr(item, "content", None)
+                if text is None and isinstance(item, dict):
+                    text = item.get("text") or item.get("content")
+            if text:
+                chunks.append(str(text))
+        joined = "".join(chunks).strip()
+        return joined or None
+    if isinstance(value, dict):
+        for key in ("content", "text", "reasoning_content", "reasoning"):
+            text = _coerce_reasoning_text(value.get(key))
+            if text:
+                return text
+        return None
+    text = str(value).strip()
+    return text or None
+
+
+def _extract_reasoning_content(obj: Any) -> str | None:
+    """Best-effort extraction across OpenAI-compatible and OCI response shapes."""
+    if obj is None:
+        return None
+
+    for attr in ("reasoning_content", "reasoning"):
+        text = _coerce_reasoning_text(getattr(obj, attr, None))
+        if text:
+            return text
+
+    if isinstance(obj, dict):
+        for key in ("reasoning_content", "reasoning"):
+            text = _coerce_reasoning_text(obj.get(key))
+            if text:
+                return text
+
+    extra = getattr(obj, "model_extra", None)
+    if isinstance(extra, dict):
+        for key in ("reasoning_content", "reasoning"):
+            text = _coerce_reasoning_text(extra.get(key))
+            if text:
+                return text
+
+    return None
+
+
 def _clean_config_value(value: Any) -> str | None:
     """Normalize values loaded from .env/YAML/PowerShell.
 
@@ -100,6 +157,9 @@ class MockLLMProvider(LLMProvider):
         self.model = "mock-llm"
 
     async def ainvoke(self, messages, **kwargs):
+        return (await self.ainvoke_response(messages, **kwargs)).content
+
+    async def ainvoke_response(self, messages, **kwargs):
         profile_name = kwargs.get("profile_name", "default")
         component_name = kwargs.get("component_name") or kwargs.get("component") or profile_name or "default"
         generation_name = kwargs.get("generation_name") or f"llm.{component_name}"
@@ -125,7 +185,15 @@ class MockLLMProvider(LLMProvider):
             generation.set_metadata(**usage)
         if self.usage_repository:
             await self.usage_repository.record(UsageRecord.from_usage("mock", model, generation_name, usage, llm_metadata))
-        return answer
+        return LLMResponse(
+            content=answer,
+            reasoning_content=None,
+            provider="mock",
+            model=model,
+            profile_name=profile_name,
+            usage=dict(usage),
+            metadata=dict(llm_metadata),
+        )
 
 
 class OCICompatibleOpenAIProvider(LLMProvider):
@@ -208,6 +276,9 @@ class OCICompatibleOpenAIProvider(LLMProvider):
         return self._clients[key]
 
     async def ainvoke(self, messages, **kwargs):
+        return (await self.ainvoke_response(messages, **kwargs)).content
+
+    async def ainvoke_response(self, messages, **kwargs):
         profile_name = kwargs.pop("profile_name", None)
         component_name = kwargs.pop("component_name", None) or kwargs.pop("component", None) or profile_name or "default"
         generation_name = kwargs.pop("generation_name", None) or f"llm.{component_name}"
@@ -227,7 +298,7 @@ class OCICompatibleOpenAIProvider(LLMProvider):
 
         if provider == "mock":
             mock = MockLLMProvider(self.settings, telemetry=self.telemetry, usage_repository=self.usage_repository)
-            return await mock.ainvoke(
+            return await mock.ainvoke_response(
                 messages,
                 model=model,
                 profile_name=resolved_profile_name,
@@ -241,7 +312,7 @@ class OCICompatibleOpenAIProvider(LLMProvider):
 
         if provider == "oci_sdk":
             sdk = OCISDKProvider(self.settings, telemetry=self.telemetry, usage_repository=self.usage_repository)
-            return await sdk.ainvoke(
+            return await sdk.ainvoke_response(
                 messages,
                 model=model,
                 temperature=temperature,
@@ -337,7 +408,9 @@ class OCICompatibleOpenAIProvider(LLMProvider):
                     model_parameters=model_parameters,
                 ) as generation:
                     resp = await client.chat.completions.create(**request_kwargs)
-                    answer = resp.choices[0].message.content or ""
+                    message = resp.choices[0].message
+                    answer = message.content or ""
+                    reasoning_content = _extract_reasoning_content(message)
 
                     usage_metadata = self.token_collector.enrich(model, getattr(resp, "usage", None))
                     usage_metadata.update({
@@ -358,7 +431,15 @@ class OCICompatibleOpenAIProvider(LLMProvider):
                         UsageRecord.from_usage(provider, model, generation_name, usage_metadata, llm_metadata)
                     )
 
-                return answer
+                return LLMResponse(
+                    content=answer,
+                    reasoning_content=reasoning_content,
+                    provider=provider,
+                    model=model,
+                    profile_name=resolved_profile_name,
+                    usage=dict(usage_metadata),
+                    metadata=dict(llm_metadata),
+                )
             except Exception as exc:
                 logger.exception(
                     "Erro ao chamar LLM provider=%s component=%s profile=%s model=%s: %s",
@@ -576,7 +657,26 @@ class OCISDKProvider(LLMProvider):
 
         return str(chat_response)
 
+    @staticmethod
+    def _extract_reasoning_content(response) -> str | None:
+        data = getattr(response, "data", response)
+        chat_response = getattr(data, "chat_response", None) or data
+
+        text = _extract_reasoning_content(chat_response)
+        if text:
+            return text
+
+        choices = getattr(chat_response, "choices", None) or []
+        if choices:
+            first = choices[0]
+            message = getattr(first, "message", None)
+            return _extract_reasoning_content(message) or _extract_reasoning_content(first)
+        return None
+
     async def ainvoke(self, messages, **kwargs):
+        return (await self.ainvoke_response(messages, **kwargs)).content
+
+    async def ainvoke_response(self, messages, **kwargs):
         import asyncio
 
         model = _clean_config_value(kwargs.get("model") or self.model)
@@ -667,6 +767,7 @@ class OCISDKProvider(LLMProvider):
             ) as generation:
                 response = await asyncio.to_thread(client.chat, details)
                 answer = self._extract_answer(response)
+                reasoning_content = self._extract_reasoning_content(response)
 
                 usage_metadata = {
                     "prompt_tokens": max(1, len(str(messages)) // 4),
@@ -693,7 +794,15 @@ class OCISDKProvider(LLMProvider):
                     )
                 )
 
-            return answer
+            return LLMResponse(
+                content=answer,
+                reasoning_content=reasoning_content,
+                provider="oci_sdk",
+                model=model,
+                profile_name=profile_name,
+                usage=dict(usage_metadata),
+                metadata=dict(llm_metadata),
+            )
 
 
 def create_llm(settings, telemetry=None, usage_repository: UsageRepository | None = None) -> LLMProvider:

libs/agent_framework/src/agent_framework/llm/types.py

@@ -0,0 +1,22 @@
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from typing import Any
+
+
+@dataclass(slots=True)
+class LLMResponse:
+    """Canonical rich response returned by LLM providers.
+
+    ``content`` preserves the legacy textual answer. ``reasoning_content`` is
+    optional because not every model/provider/API exposes reasoning text.
+    Consumers must never depend on it being present.
+    """
+
+    content: str
+    reasoning_content: str | None = None
+    provider: str | None = None
+    model: str | None = None
+    profile_name: str | None = None
+    usage: dict[str, Any] = field(default_factory=dict)
+    metadata: dict[str, Any] = field(default_factory=dict)

tests/unit/test_llm_rich_response.py

@@ -0,0 +1,66 @@
+from types import SimpleNamespace
+
+import pytest
+
+from agent_framework.llm.base import LLMProvider
+from agent_framework.llm.providers import MockLLMProvider, OCISDKProvider, _extract_reasoning_content
+from agent_framework.llm.types import LLMResponse
+
+
+class LegacyOnlyProvider(LLMProvider):
+    async def ainvoke(self, messages, **kwargs) -> str:
+        return "legacy-answer"
+
+
+@pytest.mark.asyncio
+async def test_legacy_provider_gets_rich_response_fallback_without_breaking_contract():
+    provider = LegacyOnlyProvider()
+
+    legacy = await provider.ainvoke([{"role": "user", "content": "hello"}])
+    rich = await provider.ainvoke_response([{"role": "user", "content": "hello"}])
+
+    assert legacy == "legacy-answer"
+    assert isinstance(legacy, str)
+    assert isinstance(rich, LLMResponse)
+    assert rich.content == legacy
+    assert rich.reasoning_content is None
+
+
+@pytest.mark.asyncio
+async def test_mock_ainvoke_remains_string_and_rich_api_is_opt_in():
+    provider = MockLLMProvider()
+    messages = [{"role": "user", "content": "hello"}]
+
+    legacy = await provider.ainvoke(messages)
+    rich = await provider.ainvoke_response(messages)
+
+    assert isinstance(legacy, str)
+    assert legacy == rich.content
+    assert rich.provider == "mock"
+    assert rich.model == "mock-llm"
+    assert rich.reasoning_content is None
+    assert rich.usage["total_tokens"] > 0
+
+
+def test_openai_compatible_reasoning_content_attribute_is_extracted():
+    message = SimpleNamespace(content="answer", reasoning_content="model reasoning")
+    assert _extract_reasoning_content(message) == "model reasoning"
+
+
+def test_openai_compatible_reasoning_content_model_extra_is_extracted():
+    message = SimpleNamespace(content="answer", model_extra={"reasoning_content": "extra reasoning"})
+    assert _extract_reasoning_content(message) == "extra reasoning"
+
+
+def test_missing_reasoning_content_is_none():
+    message = SimpleNamespace(content="answer")
+    assert _extract_reasoning_content(message) is None
+
+
+def test_oci_sdk_reasoning_is_extracted_from_choice_message():
+    message = SimpleNamespace(content="answer", reasoning_content="oci reasoning")
+    choice = SimpleNamespace(message=message)
+    chat_response = SimpleNamespace(choices=[choice])
+    response = SimpleNamespace(data=SimpleNamespace(chat_response=chat_response))
+
+    assert OCISDKProvider._extract_reasoning_content(response) == "oci reasoning"