Configuration
configure()
Call configure() before any Bloomberg request to set connection parameters, pool sizes, and auth. Once the engine starts on the first request, calling configure() again raises RuntimeError.
import xbbg
from xbbg import blp, EngineConfig
# Keyword arguments — most common
xbbg.configure(request_pool_size=4, subscription_pool_size=2)
# EngineConfig object
xbbg.configure(EngineConfig(request_pool_size=4, subscription_pool_size=2))
# EngineConfig + overrides (kwargs take precedence)
cfg = EngineConfig(request_pool_size=4)
xbbg.configure(cfg, subscription_pool_size=2)configure() is also available on the blp module:
from xbbg import blp
blp.configure(host='192.168.1.100', port=18194)WARNING
configure() must be called before the first Bloomberg request. Calling it after the engine has started raises RuntimeError: Cannot configure after engine has started.
Engine Architecture
xbbg runs a Rust async engine with pre-warmed worker pools:
┌─────────────────────────────────────────────────┐
│ xbbg Engine │
│ │
│ ┌──────────────────────┐ ┌─────────────────┐ │
│ │ Request Worker Pool │ │ Subscription │ │
│ │ (request_pool_size) │ │ Session Pool │ │
│ │ │ │ (sub_pool_size) │ │
│ │ Worker 1 ──session │ │ │ │
│ │ Worker 2 ──session │ │ Session 1 │ │
│ │ ... │ │ ... │ │
│ └──────────────────────┘ └─────────────────┘ │
│ │ round-robin │ isolated │
│ ▼ ▼ │
│ bdp/bdh/bds/bdib subscribe/stream │
│ bql/bsrch/beqs vwap/mktbar/depth │
└─────────────────────────────────────────────────┘- Request workers each hold an independent Bloomberg session. Concurrent
bdp/bdh/bdscalls are dispatched round-robin.request_pool_size=4allows 4 parallel Bloomberg requests. - Subscription sessions are isolated per session to prevent cross-contamination between topic streams. Each
subscribe()call gets its own session from the pool. - Workers are pre-warmed at first use — sessions are started and services opened before your first request, eliminating cold-start latency.
EngineConfig Reference
Connection & Session
| Parameter | Default | Description |
|---|---|---|
host | 'localhost' | Bloomberg server host for direct Terminal/B-PIPE TCP connections |
port | 8194 | Bloomberg server port for direct Terminal/B-PIPE TCP connections |
servers | [] | Ordered (host, port) failover list for direct TCP connections; do not combine with zfp_remote |
zfp_remote | None | Bloomberg ZFP leased-line remote, either '8194' or '8196'; mutually exclusive with host/port/servers/SOCKS5 and requires TLS credentials |
num_start_attempts | 3 | Retries before giving up on session start |
auto_restart_on_disconnection | True | Auto-reconnect on session disconnect |
Worker Pools
| Parameter | Default | Description |
|---|---|---|
request_pool_size | 2 | Number of pre-warmed request workers (parallel Bloomberg sessions for bdp/bdh/bds/etc.) |
subscription_pool_size | 1 | Number of pre-warmed subscription sessions (isolated sessions for subscribe/stream) |
warmup_services | ['//blp/refdata', '//blp/apiflds'] | Services to pre-open on startup |
Subscription Tuning
| Parameter | Default | Description |
|---|---|---|
subscription_flush_threshold | 1 | Ticks buffered before flushing to Python. Increase for throughput; decrease for latency |
subscription_stream_capacity | 256 | Backpressure buffer size per subscription stream |
overflow_policy | 'drop_newest' | Slow consumer policy: 'drop_newest' or 'block' |
Internal Buffers
| Parameter | Default | Description |
|---|---|---|
max_event_queue_size | 10000 | Bloomberg SDK event queue depth |
command_queue_size | 256 | Internal command channel capacity |
Validation
| Parameter | Default | Description |
|---|---|---|
validation_mode | 'disabled' | Field validation: 'disabled', 'strict' (reject unknown fields), or 'lenient' (warn) |
field_cache_path | ~/.xbbg/field_cache.json | Path for persistent field type cache |
Authentication (SAPI / B-PIPE)
| Parameter | Default | Description |
|---|---|---|
auth_method | None | Auth mode: 'user', 'app', 'userapp', 'dir', 'manual', or 'token' |
app_name | None | Application name (required for app, userapp, manual) |
user_id | None | Bloomberg user ID (required for manual) |
ip_address | None | Bloomberg IP address (required for manual) |
dir_property | None | Active Directory property (required for dir) |
token | None | Auth token (required for token) |
Authentication Examples
B-PIPE app
from xbbg import blp
# Application auth — requires app name registered with Bloomberg
blp.configure(
auth_method='app',
app_name='myapp:8888',
host='bpipe-host',
port=8194,
)manual (SAPI)
from xbbg import blp
# Manual auth — provides explicit Bloomberg user and IP
blp.configure(
auth_method='manual',
app_name='myapp',
user_id='12345',
ip_address='10.0.0.1',
)dir
from xbbg import blp
# Active Directory auth — uses an AD property to identify the user
blp.configure(
auth_method='dir',
dir_property='mail',
)token
from xbbg import blp
# Token auth
blp.configure(
auth_method='token',
token='<your-auth-token>',
)Multi-Server Failover
Pass a list of (host, port) tuples to servers. When set, servers overrides host/port.
from xbbg import blp
blp.configure(
servers=[
('bbg-primary.example.com', 8194),
('bbg-secondary.example.com', 8194),
]
)Bloomberg will attempt servers in order and fall back on failure.
ZFP over Leased Lines
For Bloomberg Zero-Footprint (ZFP) connectivity over private leased lines, set zfp_remote to the remote string Bloomberg assigned for the leased-line path. Accepted values are '8194' and '8196'. xbbg delegates endpoint discovery to Bloomberg's ZfpUtil::getOptionsForLeasedLines, so you must not pass direct endpoint settings in the same configuration.
ZFP mode requires both TLS files: tls_client_credentials (client PKCS#12 file) and tls_trust_material (trust material file). The password is optional only if your client credentials file is not password-protected.
from xbbg import blp
blp.configure(
zfp_remote='8194',
tls_client_credentials='/secure/client.p12',
tls_client_credentials_password='secret',
tls_trust_material='/secure/trust.p7',
)Do not combine zfp_remote with host, port, servers, socks5_host, or socks5_port. In ZFP mode those endpoint addresses are supplied by Bloomberg; passing direct endpoints now raises ValueError instead of silently trying localhost:8194.
To verify the session start path, enable SDK tracing before configure() and confirm the first transport lines contain Bloomberg infrastructure endpoints, not localhost or 127.0.0.1:
import xbbg
from xbbg import blp
xbbg.set_log_level('trace')
xbbg.enable_sdk_logging('trace')
blp.configure(
zfp_remote='8194',
tls_client_credentials='/secure/client.p12',
tls_client_credentials_password='secret',
tls_trust_material='/secure/trust.p7',
)
print(xbbg.is_connected())TLS
| Parameter | Default | Description |
|---|---|---|
tls_client_credentials | None | Path to client certificate file (PKCS#12) |
tls_client_credentials_password | None | Password for the client certificate |
tls_trust_material | None | Path to trust store file (PKCS#12) |
tls_handshake_timeout_ms | None | TLS handshake timeout in milliseconds |
tls_crl_fetch_timeout_ms | None | CRL fetch timeout in milliseconds |
from xbbg import blp
blp.configure(
host='bbg-tls.example.com',
tls_client_credentials='/certs/client.p12',
tls_client_credentials_password='changeit',
tls_trust_material='/certs/trust.p12',
tls_handshake_timeout_ms=5000,
)SOCKS5 Proxy
Route Bloomberg connections through a SOCKS5 proxy by setting both socks5_host and socks5_port.
| Parameter | Default | Description |
|---|---|---|
socks5_host | None | SOCKS5 proxy hostname |
socks5_port | None | SOCKS5 proxy port (required when socks5_host is set) |
from xbbg import blp
blp.configure(
host='bbg-server.example.com',
socks5_host='proxy.internal',
socks5_port=1080,
)Retry Policy
Retries apply to individual Bloomberg requests that fail transiently. Retry is disabled by default (retry_max_retries=0).
| Parameter | Default | Description |
|---|---|---|
retry_max_retries | 0 | Maximum retry attempts per request. 0 disables retry |
retry_initial_delay_ms | 1000 | Delay before first retry (milliseconds) |
retry_backoff_factor | 2.0 | Multiplier applied to delay after each retry |
retry_max_delay_ms | 30000 | Maximum delay between retries (milliseconds) |
from xbbg import blp
# Retry up to 3 times with exponential backoff: 1s → 2s → 4s (capped at 30s)
blp.configure(
retry_max_retries=3,
retry_initial_delay_ms=1000,
retry_backoff_factor=2.0,
retry_max_delay_ms=30000,
)INFO
Delay sequence: initial_delay_ms * backoff_factor^n, capped at retry_max_delay_ms. With the defaults above: 1000 ms, 2000 ms, 4000 ms.
SDK Logging
Bloomberg SDK log output is suppressed by default. Two ways to enable it:
At configure time — set sdk_log_level in EngineConfig:
from xbbg import blp
blp.configure(sdk_log_level='INFO')At runtime — call enable_sdk_logging() after configuration:
import xbbg
xbbg.enable_sdk_logging('DEBUG')Accepted levels are Bloomberg SDK level strings: 'OFF', 'FATAL', 'ERROR', 'WARN', 'INFO', 'DEBUG', 'TRACE'.
