Frequently Asked Questions about Julep Platform
julep-task-queue
agents-api
: Manages agent definitions, tasks, sessions, and orchestrates workflowsmemory-store
: Handles all data persistence including relational data and vector embeddingsintegrations-service
: Provides standardized interface for external tool usagellm-proxy
: Centralizes LLM interactions with a unified API/api/*
are routed to the Agents API service/tasks-ui/*
go to the Temporal UI service/v1/graphql
requests are directed to the Hasura serviceX-Developer-Id
headers for resource isolationUSE_BLOB_STORE_FOR_TEMPORAL
is enabled. It integrates with S3-compatible storage through environment variables:
S3_ENDPOINT
, S3_ACCESS_KEY
, and S3_SECRET_KEY
for connectionBLOB_STORE_BUCKET
defines the bucket nameBLOB_STORE_CUTOFF_KB
sets the size threshold for blob storagelitellm-config.yaml
with model names, parameters, and API keysfinish_reason
from βeosβ to βstopβ)GUNICORN_WORKERS
TEMPORAL_MAX_CONCURRENT_ACTIVITIES
POOL_MAX_SIZE
developer_id
for multi-tenancy.
embeddings
field stored in docs_embeddings_store
tablesearch_docs_by_embedding
for semantic similaritysearch_docs_by_text
using PostgreSQL full-text searchsearch_docs_hybrid
combining both approachescreated_at
, updated_at
) for temporal datalist_entries
with date ranges)string
or array[string]
in the Agent entityname
, description
, input_schema
, main
(workflow steps), and tools
created_at
, updated_at
) provide implicit version trackingid
, user
, agent
, situation
(context), system_template
, and metadata
id
, session_id
, role
(user/assistant/system), content
(string or JSON), source
, and timestamp
developer_id
filteringtrigram_similarity_threshold
parameter)handle_step
method to process each step typeeval_step_exprs
queued
, starting
, running
, succeeded
, failed
, cancelled
init
, step
, finish
, error
, cancelled
output
, current
, and next
workflow stepsexecution_transitions
tablecreate_execution_transition
function records all state changeserror
field of the Execution object is populatedIfElseWorkflowStep
: Evaluates condition and executes βthenβ or βelseβ branchSwitchStep
: Multi-way branching with multiple casesForeachStep
: Iterates over collections processing each itemMapReduceStep
: Maps functions over collections with optional parallel execution_handle_IfElseWorkflowStep
and _handle_ForeachStep
.
DEFAULT_RETRY_POLICY
applied to retryable errorslast_error
attribute stores recent errorsschedule_to_close_timeout
and heartbeat_timeout
for activitiesRemoteObject
pattern to optimize memory/agents
: Create, retrieve, update, delete agent definitions/tasks
: Define and execute tasks, retrieve workflow definitions/sessions
: Manage conversational sessions and conversation history/executions
: Track task executions and monitor status/docs
: Handle document storage, search, and retrieval with embeddings/tools
: Define and manage agent tools/users
: Manage user accounts and authentication/responses
: OpenAI-compatible interface for LLM responsesid
, agent_id
, user_id
, created_at
, situation contextjulep.sessions.create
linking agent and userjulep.sessions.chat
with role and contentcreate_julep_session
, session_chat
)Doc
entities) provide agent knowledge:
docs_embeddings_store
AGENTS_API_KEY
for authenticationX-Auth-Key
header requiredX-Developer-Id
header for resource isolationdefault_system_template
max_free_sessions
and max_free_executions
environment variables definedstream
configuration option availablelitellm-config.yaml
StepContext
and appropriate definitionsToolCallStep
defines tool and argumentsvalidate_py_expression
function validates expressions$
, _
, or containing {{
_
, inputs
, outputs
, state
, steps
setup
parameters of tool definitionsintegration
provider
, method
, and setup
parametersperform_browser_action
toolTEMPORAL_MAX_CONCURRENT_ACTIVITIES
controls concurrencyTEMPORAL_MAX_ACTIVITIES_PER_SECOND
limits activity rateGUNICORN_WORKERS
for integration service scalingagents-api
: Watches ./agents_api
and gunicorn_conf.py
worker
: Watches ./agents_api
and Dockerfile.worker
integrations
: Watches its directory for changesmkdir julep-responses-api
.env
filedocker compose up --watch
docker ps
npx
or uvx
.
SKIP_CHECK_DEVELOPER_HEADERS=True
AGENTS_API_KEY
AGENTS_API_MULTI_TENANT_MODE: true
models.tsp
)scripts/generate_openapi_code.sh
generates codeunittest.mock.patch
limit
and offset
parameterssearch_window
for time-based filtering (default 4 weeks)connection_lifetime
: 600 secondsidle_timeout
: 180 secondsmax_connections
: 50retries
: 1use_prepared_statements
: truePOOL_MAX_SIZE
configurable (default: CPU count, max 10)developer_id
X-Developer-Id
header for request routingX-Developer-Id
header enforcementAGENTS_API_KEY
and JWT_SHARED_KEY
for authWaitForInputStep
provides pause mechanismprevious_response_id
links responsesintegration
@function_tool
decorator for functionsIfElseWorkflowStep
for conditional logicSwitchStep
for complex branchingWaitForInputStep
for user inputErrorWorkflowStep
for invalid states$
prefix for consistency