API reference¶
Generated reference documentation for each package. The pages are rebuilt and published with this
site from main.
| Language | Reference | Generated by |
|---|---|---|
| TypeScript | TypeScript API | TypeDoc |
| Rust | Rust API · docs.rs | rustdoc |
| Go | pkg.go.dev | Go doc comments |
| Python | below | mkdocstrings |
| Java | Java API | Javadoc |
Build them locally with make docs (per language: make -C <lang> docs).
Python¶
Grow Codes SDK for SigNoz.
from grow_signoz import SignozClient, SignozConfig client = SignozClient.from_env() # doctest: +SKIP
AttrValue
module-attribute
¶
Attribute value: scalar, list of values or nested mapping (OTel AnyValue without bytes).
AggregateBuilder ¶
Bases: _AggregateSpec
Fluent builder for SigNoz aggregations (requestType scalar or time_series).
client.query.aggregate("traces").count("calls").p99("duration_nano", "p99")
.group_by("name").run(). Aliases are mapped back from SigNoz's __result_<i> columns
by aggregation index. Default window: the last hour.
AggregatePoint
dataclass
¶
One point of a time series.
AggregateResult
dataclass
¶
Result of :meth:AggregateBuilder.run.
AggregateRow
dataclass
¶
One scalar result row.
AggregateSeries
dataclass
¶
One series of :meth:AggregateBuilder.time_series.
Aggregation
dataclass
¶
One aggregation: a SigNoz expression such as count() or p99(duration_nano).
ApiResult
dataclass
¶
Decoded payload (data of the envelope) and the envelope's warning messages.
AsyncAggregateBuilder ¶
AsyncBaseLogQueryClient ¶
Async :class:BaseLogQueryClient.
AsyncBaseTraceQueryClient ¶
Async :class:BaseTraceQueryClient.
query_trace_id ¶
The id wait_for_trace passes to get_trace; override to keep native ids.
AsyncBatchProcessor ¶
Bases: _BatchCore
Batches records for an :class:AsyncLogExporter on an asyncio task.
The task starts on the running loop at the first :meth:add (or :meth:flush). :meth:add
may be called from other threads; it wakes the worker thread-safely. Exports are bounded by
export_timeout_s with :func:asyncio.wait_for.
add ¶
Queue without blocking; False when dropped (block acts as drop_newest here).
add_wait
async
¶
Queue a record, waiting for queue space with drop_policy="block".
flush
async
¶
Export everything queued now; raises TimeoutError when timeout_s passes first.
shutdown
async
¶
Drain (bounded by timeout_s), stop the task, shut the exporter down; idempotent.
AsyncBatchingLogClient ¶
Bases: _ClientBase
:class:AsyncLogClient over an :class:AsyncLogExporter + :class:AsyncBatchProcessor.
AsyncHttpTransport ¶
Asynchronous transport (one per async client); options as for :class:HttpTransport.
Cancelling the calling task raises :class:CancelledError (an asyncio.CancelledError
subclass) so asyncio's cancellation machinery keeps working. Every call after
:meth:aclose raises :class:CancelledError ("client is closed").
AsyncLogClient ¶
Bases: Protocol
Asynchronous log client: send stays synchronous, flush/close are awaited.
AsyncLogExporter ¶
Bases: Protocol
Asynchronous exporter, awaited by :class:AsyncBatchProcessor on the caller's event loop.
AsyncLogQueryClient ¶
Bases: Protocol
Asynchronous query API (same methods, awaited).
AsyncPollingTail ¶
Async :class:PollingTail; the polling task runs on the current event loop.
AsyncSignozClient ¶
Asynchronous SigNoz client; use async with or call :meth:aclose.
Same surface as :class:SignozClient with awaited methods (:attr:query,
:attr:traces, :attr:logs, :meth:log_client returns an
:class:AsyncBatchingLogClient). Cancel a call by cancelling its task: the SDK raises
:class:CancelledError, which is also an :class:asyncio.CancelledError.
resources
property
¶
Synchronous background resources stopped by :meth:aclose.
traces
property
¶
Tracing-category query client (the same object as :attr:query).
aclose
async
¶
Flush and close log clients, tracing handles and tails, then close connections.
Idempotent.
from_env
classmethod
¶
Build a client from GROW_SIGNOZ_* and vendor environment variables.
init_tracing ¶
See :meth:SignozClient.init_tracing; shut down by :meth:aclose.
log_client ¶
log_client(*, service_name: str | None = None, service_version: str | None = None, environment: str | None = None, protocol: OtlpProtocol | None = None, endpoint: str | None = None, headers: Mapping[str, str] | None = None, options: BatchOptions | None = None, resource: Mapping[str, AttrValue] | None = None, redaction: Any = _UNSET, correlate: bool = True) -> AsyncBatchingLogClient
An async OTLP log client (see :func:create_async_log_client), closed by
:meth:aclose. Raises :class:CancelledError when the client is closed.
request
async
¶
Send a raw request through the client's transport.
AsyncSignozLogExporter ¶
Bases: _ExporterBase
AsyncLogExporter variant of :class:SignozLogExporter (gRPC runs in a thread).
AsyncSignozLogQueryClient ¶
Bases: AsyncBaseLogQueryClient
Async :class:SignozLogQueryClient; get it from :attr:AsyncSignozClient.logs.
search
async
¶
search(query: str, *, start: datetime | None = None, end: datetime | None = None, limit: int = 100, order: SearchOrder = 'desc', cursor: str | None = None) -> LogSearchResult
See :meth:SignozLogQueryClient.search.
tail
async
¶
tail(query: str, on_record: Callable[[LogRecord], None], *, interval_s: float | timedelta = 1.0, start: datetime | None = None, on_error: Callable[[Exception], None] | None = None, reconnect_delay_s: float | timedelta | None = None) -> AsyncSignozTailHandle
See :meth:SignozLogQueryClient.tail (runs as a task on the current loop).
AsyncSignozQueryClient ¶
Bases: AsyncBaseTraceQueryClient
Async :class:SignozQueryClient; get it from :attr:AsyncSignozClient.query.
aggregate ¶
Start an async aggregation over logs or traces.
get_dependencies
async
¶
get_dependencies(*, start: datetime | None = None, end: datetime | None = None) -> list[DependencyLink]
See :meth:SignozQueryClient.get_dependencies.
get_trace
async
¶
See :meth:SignozQueryClient.get_trace.
list_field_values
async
¶
list_field_values(name: str, *, source: SignozSignal = 'logs', search: str | None = None) -> list[str | int | float | bool]
See :meth:SignozQueryClient.list_field_values.
list_log_fields
async
¶
See :meth:SignozQueryClient.list_log_fields.
list_operations
async
¶
list_operations(service: str, *, start: datetime | None = None, end: datetime | None = None) -> list[str]
See :meth:SignozQueryClient.list_operations.
list_services
async
¶
See :meth:SignozQueryClient.list_services.
query_logs
async
¶
query_logs(filter: str | FilterBuilder | None = None, *, start: datetime | None = None, end: datetime | None = None, limit: int | None = None, order_by: SortOrder = 'desc', cursor: str | None = None, select_fields: Sequence[str] | None = None) -> LogQueryResult
See :meth:SignozQueryClient.query_logs.
query_range
async
¶
Post a raw /api/v5/query_range body; returns data plus warnings.
query_traces
async
¶
query_traces(*, service: str | None = None, operation: str | None = None, filter: str | FilterBuilder | None = None, start: datetime | None = None, end: datetime | None = None, limit: int | None = None, offset: int = 0, order_by: SortOrder = 'desc') -> SpanQueryResult
See :meth:SignozQueryClient.query_traces.
search_traces
async
¶
search_traces(query: TraceSearchQuery | None = None, *, filter: str | FilterBuilder | None = None) -> list[TraceSummary]
See :meth:SignozQueryClient.search_traces.
AsyncSignozTailHandle ¶
A running async live tail (task on the current event loop); await stop().
AsyncTraceQueryClient ¶
Bases: Protocol
Asynchronous trace query API (same methods, awaited).
AuthenticationError ¶
Bases: SignozError
Credentials missing or rejected (HTTP 401).
BaseConfig
dataclass
¶
Common, immutable client settings. Durations are float seconds (timedelta accepted).
Service configs subclass this (@dataclass(frozen=True, repr=False)), override url
with the service default, add their own fields (Secret for credentials), extend
:meth:_env_values and :meth:validate, and list vendor variables in ENV_VENDOR.
BaseLogQueryClient ¶
Base for service query clients: implement :meth:search, override what the service has.
tail defaults to :class:PollingTail; list_streams/list_labels raise
UnsupportedError; wait_for_log uses :func:wait_for_log.
BaseTraceQueryClient ¶
Base for service query clients; every method raises UnsupportedError until overridden
(wait_for_trace uses :func:wait_for_trace on top of get_trace, passing the id
through :meth:query_trace_id).
query_trace_id ¶
The id wait_for_trace passes to get_trace; override to keep native ids.
BatchOptions
dataclass
¶
Batching options; durations are float seconds (timedelta accepted).
max_export_retries/retry_backoff_s: a batch whose export failed with a retryable
SDK error (raised, or returned with nothing accepted) is exported again up to
max_export_retries times, waiting retry_backoff_s * 2**n before retry n.
Partial acceptance is final.
from_mapping
classmethod
¶
Build from the camelCase form used by conformance scenarios (flushIntervalMs …).
BatchProcessor ¶
Bases: _BatchCore
Batches records for a synchronous :class:LogExporter on a daemon worker thread.
The worker starts with the first record. It exports when a batch is full, when
flush_interval_s passed since the first queued record, on :meth:flush and on
:meth:shutdown. Processors with a started worker are shut down at interpreter exit.
add ¶
Queue a record; False when it was dropped. Blocks only with drop_policy="block".
flush ¶
Export everything queued now; raises TimeoutError when timeout_s passes first.
shutdown ¶
Drain the queue (bounded by timeout_s), stop the worker and shut the exporter down.
Idempotent. Records still queued when the timeout passes are dropped with reason
shutdown; later :meth:add calls are dropped the same way.
BatchStats
dataclass
¶
Counters of a batch processor (see the module docs for their meaning).
BatchingLogClient ¶
Bases: _ClientBase
:class:LogClient over a synchronous :class:LogExporter and a :class:BatchProcessor.
redaction defaults to RedactionProcessor(); pass None to disable it.
correlate adds the active span's ids to records without a trace id.
CancelToken ¶
Thread-safe cancellation flag passed to sync calls as cancel=.
Cancelling interrupts an in-flight request (the call returns immediately with
:class:CancelledError; the abandoned HTTP exchange finishes in the background) and any
retry backoff sleep.
CancelledError ¶
Bases: SignozError, CancelledError
The caller cancelled the operation (task cancel, :class:CancelToken, client close).
Also an :class:asyncio.CancelledError so asyncio keeps treating the task as cancelled.
ConfigError ¶
Bases: SignozError
Invalid or missing configuration.
ConflictError ¶
Bases: SignozError
Conflict or failed precondition (HTTP 409, 412, CAS mismatch).
EnvReader ¶
Reads GROW_<KEY_UPPER>_<NAME> and vendor variables from an injectable mapping.
Every getter returns None when no variable is set (empty strings count as unset). A
present but invalid value raises :class:ConfigError that names the variable, never its
value.
ErrorCode ¶
Bases: str, Enum
Canonical error codes; the value is the wire form used by the conformance suite.
ErrorInfo
dataclass
¶
Passed to Hooks.on_error; error is the (redacted) SDK error.
ExportResult
dataclass
¶
Outcome of one :meth:LogExporter.export call.
retries is a Python extension: the number of transport retries the export needed
(summed into BatchStats.retries).
FilterBuilder ¶
Immutable builder for SigNoz filter expressions.
FilterBuilder().eq("service.name", "x").in_("severity_text", ["WARN", "ERROR"]).build()
→ service.name = 'x' AND severity_text IN ('WARN', 'ERROR'). Clauses are joined with
AND; use :meth:or_ / :meth:not_ for groups (rendered in parentheses). Every method
returns a new builder; invalid field names, empty lists and non-finite numbers raise
:class:ValidationError.
contains ¶
field CONTAINS 'text' — matches text literally (LIKE wildcards escaped).
in_ ¶
field IN (v1, v2, …) (non-empty list).
like ¶
field LIKE 'pattern' — % and _ are wildcards; passed as given.
not_contains ¶
NOT field CONTAINS 'text' (literal match).
not_in ¶
field NOT IN (v1, v2, …) (non-empty list).
or_ ¶
((a) OR (b) …) from builders/expressions; empty alternatives are ignored.
raw ¶
A raw expression, passed through untouched (parenthesised when it contains OR).
GroupByField
dataclass
¶
A group-by field with an optional context (resource/attribute/span/log).
HealthStatus
dataclass
¶
Result of health(); status is ok/degraded/unhealthy/unreachable or a vendor word.
Hooks
dataclass
¶
Optional callbacks invoked on every attempt. Exceptions raised by hooks are ignored.
HttpTransport ¶
Synchronous transport (one per client).
Service options (all optional):
request_id_headers: vendor request id headers, checked before :data:REQUEST_ID_HEADERS.retry_after_headers: headers read (in order) for the delay a 429/503 response asks for; default("retry-after",), e.g.("retry-after", "x-ratelimit-reset"). See :func:retry_after_from_headers.error_mapper:mapper(error, response)returns a service-specific error for an HTTP error response (orNoneto keep the default). The mapped error is whatHooks.on_error, the retry planner (retry_after) and the caller see; whether a status is retried stays the SPEC rule. Exceptions raised by the mapper are logged and ignored.
Every call after :meth:close raises :class:CancelledError ("client is closed").
LogClient ¶
Bases: Protocol
Synchronous log client (SPEC LogClient).
LogExporter ¶
Bases: Protocol
Synchronous exporter, called from the :class:BatchProcessor worker thread.
export sends one batch and reports how many records the service accepted. It may raise
an SDK error instead of returning ExportResult(error=...); both count as failed records.
timeout_s is the per-batch export budget the exporter should pass to its transport.
LogField
dataclass
¶
A log field known to SigNoz (one entry per name and context).
LogHandler ¶
Bases: Handler
:class:logging.Handler that forwards records to a log client.
Attributes: extra= fields, logger.name, code.filepath/code.function/
code.lineno, thread.name and exception.* for exc_info. Records from loggers
in exclude_loggers (prefix match; defaults to the HTTP stack the exporters use) and from
the SDK's own worker threads are skipped to avoid feedback loops. When a formatter is set,
the formatted text is the body; otherwise record.getMessage().
LogInput
dataclass
¶
Convenience input for :meth:LogClient.send; defaults are filled by the client.
LogQueryClient ¶
Bases: Protocol
Synchronous query API. Services without a query API raise UnsupportedError.
LogQueryResult
dataclass
¶
One page of :meth:SignozQueryClient.query_logs.
LogRecord
dataclass
¶
One log record (OTel log data model subset). Treat instances as immutable.
from_json
classmethod
¶
Parse the normalized wire form; raises ValidationError for malformed input.
to_json ¶
Normalized wire form (spec/schemas/log-record.schema.json, camelCase keys).
severity is the OTel number; timestamps are RFC 3339 UTC with microseconds.
to_json_str ¶
Compact JSON text of :meth:to_json (UTF-8 safe, non-JSON values via str).
LogSearchResult
dataclass
¶
One page of search results.
LogSink ¶
Bases: Protocol
Anything that accepts records: every log client.
Logger ¶
Small structured logger writing to a :class:LogSink.
Records below min_severity are discarded. The sink fills timestamp, resource and trace
correlation (the active OpenTelemetry span) for records without a trace id.
child ¶
Logger with the same sink and level and the extra attributes.
LoggerLike ¶
Bases: Protocol
Structured logger interface accepted in configuration (SPEC Logger).
NotFoundError ¶
Bases: SignozError
Missing record (HTTP 404).
PermissionDeniedError ¶
Bases: SignozError
Authenticated but not allowed (HTTP 403).
PollingTail ¶
tail built on search: a daemon thread polls for records newer than the last one.
Errors go to on_error (default: ignored) and polling continues. :meth:stop is idempotent.
RateLimitError ¶
Bases: SignozError
Too many requests (HTTP 429); retry_after holds the requested delay when known.
RedactionProcessor ¶
Returns redacted copies of records.
Attribute names containing one of keys (case-insensitive substring) get replacement
as value, recursively through nested mappings and lists. patterns (regular expressions or
their source strings) are replaced in the body and in every string attribute value.
:class:Secret values are always replaced.
Request
dataclass
¶
One logical API call (possibly several attempts).
path is joined to the base URL ({name} style values must already be encoded; use
:func:urllib.parse.quote); an absolute http:///https:// URL is used as is (for
pagination links the service returns; the default headers, config.headers included,
are sent there too, so only pass URLs of the same service). json is encoded when not
None; content sends raw bytes. idempotent overrides the method rule; an
idempotency_key sends the Idempotency-Key header and marks the request idempotent.
raise_for_status=False returns error responses instead of raising (retries still
apply). Statuses in allow_status are expected answers (a 404 that carries data, a
health endpoint's 503): the response is returned at once, like a success, without retries
or on_error.
RequestInfo
dataclass
¶
Passed to Hooks.on_request; url and headers are redacted.
ResourceRegistry ¶
Holds :class:Stoppable objects; :meth:stop_all stops each once, ignoring errors.
register ¶
Track item; when the registry is already closed the item is stopped now.
Response
dataclass
¶
A completed HTTP response (body fully read).
ResponseInfo
dataclass
¶
Passed to Hooks.on_response; headers are the redacted response headers (rate
limit headers and the like stay readable).
RetryPolicy
dataclass
¶
Retry settings. Durations are seconds (datetime.timedelta is accepted).
Backoff for retry n (0-based) is full jitter:
random(0, min(max_backoff, initial_backoff * multiplier ** n)).
random returns a float in [0, 1); inject it in tests for deterministic sleeps.
Secret ¶
Wraps a credential so it never shows up in str/repr/logs.
s = Secret("hunter2") str(s), s.reveal() ('[REDACTED]', 'hunter2')
Severity ¶
Bases: IntEnum
Log severity; the values are the OpenTelemetry severity numbers.
SignozClient ¶
Synchronous SigNoz client. Thread-safe; use as a context manager or call :meth:close.
- :meth:
health, :meth:ping, :meth:server_version - :attr:
query— :class:SignozQueryClient(logs, spans, traces, services, aggregations); :attr:tracesis the same object (TraceQueryClient) - :attr:
logs— :class:SignozLogQueryClient(LogQueryClient: search, tail, …) - :meth:
log_client— OTLP log client; :meth:init_tracing— OpenTelemetry tracing
:meth:close flushes and closes every log client, tracing handle and tail created through
it. Every I/O method accepts cancel= (a :class:CancelToken).
resources
property
¶
Background resources (log clients, tracing handles, tails) stopped by :meth:close.
traces
property
¶
Tracing-category query client (the same object as :attr:query).
close ¶
Flush and close log clients, tracing handles and tails, then close connections.
Idempotent.
from_env
classmethod
¶
Build a client from GROW_SIGNOZ_* and vendor environment variables.
health ¶
Unauthenticated liveness (GET /api/v1/health?live=1, also checks ClickHouse).
Never raises for an unhealthy or unreachable service (only CancelledError and
ConfigError).
init_tracing ¶
Tracing to the collector with this client's config (see :func:init_tracing); the
handle is shut down by :meth:close.
Raises :class:CancelledError when the client is closed.
log_client ¶
log_client(*, service_name: str | None = None, service_version: str | None = None, environment: str | None = None, protocol: OtlpProtocol | None = None, endpoint: str | None = None, headers: Mapping[str, str] | None = None, options: BatchOptions | None = None, resource: Mapping[str, AttrValue] | None = None, redaction: Any = _UNSET, correlate: bool = True) -> BatchingLogClient
An OTLP log client for the collector (see :func:create_log_client); it is
flushed and closed by :meth:close.
Raises :class:CancelledError when the client is closed.
ping ¶
Authenticated round trip (GET /api/v1/fields/keys with SIGNOZ-API-KEY).
Returns the latency in milliseconds; :class:AuthenticationError on a bad key,
:class:ConfigError without one.
request ¶
Send a raw request through the client's transport (retries, hooks, redaction).
server_version ¶
Server version from GET /api/v1/version (leading v stripped).
The official community image reports an empty version → :class:UnsupportedError.
SignozConfig
dataclass
¶
Bases: BaseConfig
Immutable settings for :class:~grow_signoz.SignozClient.
Common fields (url, timeouts, retry, tls, proxy, headers,
user_agent_suffix, hooks, logger) come from :class:BaseConfig. Durations are
float seconds; datetime.timedelta is accepted. api_key and ingestion_key accept a
str and are stored as :class:Secret; repr() never shows them.
Service fields:
api_key— query API key sent asSIGNOZ-API-KEY. Needed only for query calls and :meth:~grow_signoz.SignozClient.ping; an ingest-only client does not need it.otlp_endpoint— collector endpoint (defaulthttp://127.0.0.1:4318forhttp/protobuf,http://127.0.0.1:4317forgrpc).otlp_protocol—"http/protobuf"(default) or"grpc".ingestion_key— SigNoz Cloud ingestion key, sent assignoz-ingestion-keyon every OTLP export (self-hosted collectors ignore it).
Environment (from_env): explicit keyword > GROW_SIGNOZ_<NAME> > vendor variable >
default. URL (SIGNOZ_URL), API_KEY (SIGNOZ_API_KEY), OTLP_ENDPOINT
(SIGNOZ_OTLP_ENDPOINT, OTEL_EXPORTER_OTLP_ENDPOINT), OTLP_PROTOCOL
(OTEL_EXPORTER_OTLP_PROTOCOL), INGESTION_KEY (SIGNOZ_INGESTION_KEY), plus the
common CONNECT_TIMEOUT, REQUEST_TIMEOUT, TOTAL_TIMEOUT, MAX_RETRIES,
CA_CERT, CLIENT_CERT, CLIENT_KEY, INSECURE_SKIP_VERIFY, PROXY.
SignozLogExporter ¶
Bases: _ExporterBase
LogExporter sending OTLP logs to the SigNoz collector.
HTTP/protobuf (POST {endpoint}/v1/logs, optional gzip) or gRPC (LogsService/Export,
extra grow-signoz[grpc]). Adds signoz-ingestion-key for SigNoz Cloud when configured.
Failed batches are retried by the :class:BatchProcessor.
SignozLogQueryClient ¶
Bases: BaseLogQueryClient
LogQueryClient for SigNoz. query strings are SigNoz filter expressions; bare
words are full-text searches of the log body. Get it from :attr:SignozClient.logs.
list_labels ¶
Distinct log field names (attributes, resource attributes, intrinsic fields), sorted.
list_streams ¶
Service names that sent logs (SigNoz has no streams; service.name is the
equivalent), sorted.
search ¶
search(query: str, *, start: datetime | None = None, end: datetime | None = None, limit: int = 100, order: SearchOrder = 'desc', cursor: str | None = None) -> LogSearchResult
Search logs (default window: the last hour, newest first, 100 per page).
tail ¶
tail(query: str, on_record: Callable[[LogRecord], None], *, interval_s: float | timedelta = 1.0, start: datetime | None = None, on_error: Callable[[Exception], None] | None = None, reconnect_delay_s: float | timedelta | None = None) -> SignozTailHandle
Live tail over GET /api/v3/logs/livetail (server-sent events).
Calls on_record once per matching record from a background thread; reconnects after
the stream ends (resuming at the last seen timestamp, de-duplicating by row id) after
reconnect_delay_s (default interval_s). Authentication, permission, unsupported
and config errors stop the tail after on_error. The client's close() stops it.
SignozLogRecord
dataclass
¶
Bases: LogRecord
A :class:LogRecord read from SigNoz, with the row id and the exact timestamp.
timestamp has microsecond precision (Python datetime); time_unix_nano keeps the
exact epoch nanoseconds.
SignozQueryClient ¶
Bases: BaseTraceQueryClient
Query client for SigNoz (SIGNOZ-API-KEY auth); implements TraceQueryClient.
Logs and spans via /api/v5/query_range, traces via the waterfall API, services and
operations via v5 aggregations, dependencies via /api/v1/dependency_graph, log fields
via /api/v1/fields/keys. Get it from :attr:SignozClient.query. Every method accepts
cancel= (a :class:CancelToken).
aggregate ¶
Start an aggregation (count / percentiles / …) over logs or traces.
get_dependencies ¶
get_dependencies(*, start: datetime | None = None, end: datetime | None = None, cancel: CancelToken | None = None) -> list[DependencyLink]
Service dependency edges in the window (default: the last hour), sorted.
Edges come from client→server spans and from client spans with db.system or
messaging.system.
get_trace ¶
Full trace via POST /api/v4/traces/{id}/waterfall; None when unknown.
Span start times have millisecond precision. A trace larger than the server's waterfall window is completed with raw span queries.
list_field_values ¶
list_field_values(name: str, *, source: SignozSignal = 'logs', search: str | None = None, cancel: CancelToken | None = None) -> list[str | int | float | bool]
Distinct values of a field for logs or traces, sorted.
list_log_fields ¶
list_log_fields(*, search: str | None = None, limit: int | None = None, cancel: CancelToken | None = None) -> list[LogField]
Log fields, one entry per name and context, sorted by name then context.
list_operations ¶
list_operations(service: str, *, start: datetime | None = None, end: datetime | None = None, cancel: CancelToken | None = None) -> list[str]
Span names of service in the window (default: the last hour), sorted.
list_services ¶
list_services(*, start: datetime | None = None, end: datetime | None = None, cancel: CancelToken | None = None) -> list[str]
Every service with spans in the window (default: the last hour), sorted.
query_logs ¶
query_logs(filter: str | FilterBuilder | None = None, *, start: datetime | None = None, end: datetime | None = None, limit: int | None = None, order_by: SortOrder = 'desc', cursor: str | None = None, select_fields: Sequence[str] | None = None, cancel: CancelToken | None = None) -> LogQueryResult
Raw log query: one page of records ordered by timestamp, ties by id.
filter is a SigNoz filter expression or :class:FilterBuilder (empty: every log;
bare words search the body). Window default: the last hour (end exclusive). limit
1..10000 (None/0 → 100). Paging uses the SDK's own cursor, never the server's
nextCursor (which skips rows in the millisecond before it): the next page repeats the
boundary millisecond and excludes the ids already returned there.
query_range ¶
query_range(body: Mapping[str, Any], *, operation: str = 'query_range', cancel: CancelToken | None = None) -> ApiResult
Post a raw /api/v5/query_range body; returns data plus warnings.
query_traces ¶
query_traces(*, service: str | None = None, operation: str | None = None, filter: str | FilterBuilder | None = None, start: datetime | None = None, end: datetime | None = None, limit: int | None = None, offset: int = 0, order_by: SortOrder = 'desc', cancel: CancelToken | None = None) -> SpanQueryResult
Raw span query: one :class:Span per matching span (offset paging).
search_traces ¶
search_traces(query: TraceSearchQuery | None = None, *, filter: str | FilterBuilder | None = None, cancel: CancelToken | None = None) -> list[TraceSummary]
Trace search, newest first (default 20): a trace matches when any span matches.
start_micros, span_count (all spans), root service and name come from each
hit's waterfall (millisecond precision, 4 requests at a time); duration_micros is
the root span's.
SignozTailHandle ¶
Span
dataclass
¶
A finished span as returned by a query API.
Ids are normalized on construction (an all-zero or empty parent_span_id becomes
None); a negative duration_micros (the default) is computed from start and end.
A subclass with normalize_ids = False keeps native ids verbatim (an empty
parent_span_id still becomes None); link_class is used by :meth:from_json.
SpanBatchOptions
dataclass
¶
BatchSpanProcessor settings; durations are float seconds (timedelta accepted).
SpanLink
dataclass
¶
A link to another span. Set normalize_ids = False on a subclass for native ids.
SpanNode
dataclass
¶
A span with its children (sorted like :func:sort_spans).
SpanQueryResult
dataclass
¶
Result of :meth:SignozQueryClient.query_traces: one :class:Span per matching span.
SpanStatus
dataclass
¶
from_json
classmethod
¶
Parse {"code", "message"?} (a bare code string is accepted too).
SseParser ¶
Incremental text/event-stream parser: feed text, get (event, data) pairs.
TimeoutError ¶
Bases: SignozError
A connect, request or total deadline was exceeded (also HTTP 408, 504).
TlsConfig
dataclass
¶
TLS settings. insecure_skip_verify is for tests only and logs a warning once.
Trace
dataclass
¶
All spans of one trace, sorted by start time then span id.
A subclass with normalize_ids = False keeps a native trace id verbatim; span_class
is used by :meth:from_json.
TraceQueryClient ¶
Bases: Protocol
Synchronous trace query API.
TraceSearchQuery
dataclass
¶
matches ¶
Client-side filter: a span must match service/operation/tags; duration and time bounds apply to the whole trace.
TraceSummary
dataclass
¶
TracingASGIMiddleware ¶
ASGI middleware: one SERVER span per HTTP request, parent taken from request headers.
Non-HTTP scopes (lifespan, websocket) and excluded_paths pass through untraced.
TracingConfig
dataclass
¶
Settings for :func:init_tracing_with. repr redacts credential headers.
protocol and endpoint are interpreted by the service's exporter factory; the
template default protocol is http/protobuf. propagators beyond tracecontext,
baggage, b3 and b3multi need a factory from the service package.
protocol and propagators remember whether the caller set them: left out (or passed
as None) they read as :data:DEFAULT_TRACE_PROTOCOL / :data:DEFAULT_TRACE_PROPAGATORS
exactly as before, and :attr:protocol_set / :attr:propagators_set are False so a
service default can apply instead. The marker travels with the value, so
:meth:with_changes, dataclasses.replace and TracingConfig(**fields) keep an unset
field unset; passing the default value explicitly counts as set.
TracingHandle ¶
TracingWSGIMiddleware ¶
WSGI middleware: one SERVER span per request; ends when the response is closed.
UnavailableError ¶
Bases: SignozError
Network error or server unavailable (HTTP 500, 502, 503).
UnsupportedError ¶
Bases: SignozError
The service or deployment does not offer this capability (HTTP 501, missing API).
ValidationError ¶
Bases: SignozError
Bad input (HTTP 400, 413, 422, other 4xx, client-side validation).
add_event ¶
Add an event to the active span (no-op without one).
and_filters ¶
Join expressions with AND, parenthesising each non-empty part when there are several.
apoll_until
async
¶
apoll_until(fn: Callable[[], Awaitable[T | None]], *, timeout_s: float = 30.0, interval_s: float = 1.0, predicate: Callable[[T], bool] | None = None, operation: str = 'poll', clock: Callable[[], float] = time.monotonic) -> T
Async :func:poll_until; fn is an async callable.
async_httpx_event_hooks ¶
async_httpx_event_hooks(*, tracer: Tracer | None = None, propagator: TextMapPropagator | None = None) -> dict[str, list[Callable[..., Any]]]
event_hooks for httpx.AsyncClient (same behaviour as :func:httpx_event_hooks).
async_wait_for_log
async
¶
async_wait_for_log(client: AsyncLogQueryClient, query: str, *, timeout_s: float | timedelta = 30.0, interval_s: float | timedelta = 1.0, predicate: RecordPredicate | None = None) -> LogRecord
Async form of :func:wait_for_log.
async_wait_for_trace
async
¶
async_wait_for_trace(client: AsyncTraceQueryClient, trace_id: str, *, timeout_s: float | timedelta = 30.0, interval_s: float | timedelta = 1.0, min_spans: int = 1, normalize: Callable[[str], str] | None = normalize_trace_id) -> Trace
Async form of :func:wait_for_trace.
build_propagator ¶
build_propagator(names: Sequence[str], factories: Mapping[str, PropagatorFactory] | None = None) -> TextMapPropagator
Composite propagator for names; factories adds or overrides propagators by name.
build_resource ¶
Resource with service.name/service.version/deployment.environment and extras.
The environment is written both as deployment.environment.name (current semantic
conventions) and as the older deployment.environment. Merged on top of the SDK default
resource (telemetry.sdk.*, OTEL_RESOURCE_ATTRIBUTES); explicit config values win.
build_sampler ¶
OpenTelemetry sampler for a TracingConfig.sampler name.
build_tree ¶
Parent/child forest; spans whose parent is missing become roots.
create_async_log_client ¶
create_async_log_client(config: SignozConfig | None = None, *, service_name: str | None = None, service_version: str | None = None, environment: str | None = None, protocol: OtlpProtocol | None = None, endpoint: str | None = None, headers: Mapping[str, str] | None = None, options: BatchOptions | None = None, resource: Mapping[str, AttrValue] | None = None, redaction: Any = _UNSET, correlate: bool = True, http_transport: AsyncBaseTransport | None = None) -> AsyncBatchingLogClient
Async :func:create_log_client (an :class:AsyncBatchingLogClient).
create_log_client ¶
create_log_client(config: SignozConfig | None = None, *, service_name: str | None = None, service_version: str | None = None, environment: str | None = None, protocol: OtlpProtocol | None = None, endpoint: str | None = None, headers: Mapping[str, str] | None = None, options: BatchOptions | None = None, resource: Mapping[str, AttrValue] | None = None, redaction: Any = _UNSET, correlate: bool = True, http_transport: BaseTransport | None = None) -> BatchingLogClient
A :class:BatchingLogClient sending OTLP logs to the SigNoz collector.
Needs no API key (only ingestion_key for SigNoz Cloud). service_name (default
unknown_service), service_version and environment become resource attributes;
protocol/endpoint default to the config; options tunes the batch processor;
redaction defaults to RedactionProcessor() (None disables it).
create_signoz_span_exporter ¶
The span exporter for a config (the exporter factory of init_tracing_with).
http/protobuf → OTLP/HTTP protobuf to {endpoint}/v1/traces; grpc → OTLP/gRPC
(extra grow-signoz[grpc]). native raises :class:ConfigError.
current_span ¶
The active span (a non-recording invalid span when none is active).
current_span_id ¶
16-char lower-hex span id of the active span, or None.
current_trace_context ¶
(trace_id, span_id, trace_flags) of the active OpenTelemetry span, if any.
current_trace_id ¶
32-char lower-hex trace id of the active span, or None.
datetime_to_micros ¶
Microseconds since the epoch (naive datetimes are UTC).
default_tracer ¶
Tracer of this package from the global provider.
dependencies_from_spans ¶
Service dependency links (parent service → child service, different services only).
encode_logs ¶
Build an OTLP ExportLogsServiceRequest, grouping records by identical resource.
endpoint_for ¶
Endpoint for protocol: the configured one when the protocols match, otherwise the
configured endpoint with the standard port swapped (4318 ↔ 4317).
error_for_code ¶
Return the error class for a code (ErrorCode member or wire value).
error_from_status ¶
error_from_status(status: int, operation: str, *, content: bytes = b'', headers: Headers | None = None, known_secrets: list[str] | None = None, retry_after: float | None = None, request_id_headers: Sequence[str] = ()) -> ServiceError
Map an HTTP error status to the SPEC 2.1 error with a redacted, truncated message.
request_id_headers are checked before the default :data:REQUEST_ID_HEADERS.
escape_like_pattern ¶
Escape a value for CONTAINS/LIKE so it matches literally (\, %, _).
String quoting (:func:quote_filter_string) is applied afterwards.
extract ¶
extract(carrier: Mapping[str, str], *, context: Context | None = None, propagator: TextMapPropagator | None = None) -> otel_context.Context
Read a context from carrier (exact key match; use :func:extract_headers for HTTP).
extract_headers ¶
extract_headers(headers: HeaderSource, *, context: Context | None = None, propagator: TextMapPropagator | None = None) -> otel_context.Context
Read a context from HTTP headers (case-insensitive; mapping or ASGI-style byte pairs).
filter_literal ¶
Render a literal: numbers bare, booleans true/false, strings quoted.
format_timestamp ¶
RFC 3339 UTC form with microseconds and Z; naive datetimes are taken as UTC.
get_logger ¶
The package logger (LOGGER_NAME), silent by default (NullHandler).
grpc_error ¶
Map a gRPC status code name ("UNAVAILABLE"…) to the SPEC error classes.
httpx_event_hooks ¶
httpx_event_hooks(*, tracer: Tracer | None = None, propagator: TextMapPropagator | None = None) -> dict[str, list[Callable[..., Any]]]
event_hooks for httpx.Client: a CLIENT span per request with injected headers.
A request that fails before a response arrives (connection error) never reaches the
response hook, so its span is not exported; use :func:instrument_requests_session style
wrapping or with_span around the call when such failures must be traced.
init_tracing ¶
init_tracing(config: TracingConfig | str, *, signoz: SignozConfig | None = None, propagator_factories: Mapping[str, PropagatorFactory] | None = None, span_processors: Sequence[SpanProcessor] = (), id_generator: IdGenerator | None = None, **overrides: Any) -> TracingHandle
Set up OpenTelemetry tracing that exports to the SigNoz collector (SPEC §5).
config is a :class:TracingConfig or the service name (**overrides are then
TracingConfig fields). Endpoint, protocol default and ingestion key come from signoz
(default :meth:SignozConfig.from_env); explicit TracingConfig values win. Raises
:class:ConfigError for invalid settings.
init_tracing_with ¶
init_tracing_with(config: TracingConfig, exporter_factory: ExporterFactory, *, propagator_factories: Mapping[str, PropagatorFactory] | None = None, span_processors: Sequence[SpanProcessor] = (), id_generator: IdGenerator | None = None, sampler: Sampler | None = None) -> TracingHandle
Build a TracerProvider for config with a BatchSpanProcessor around the exporter.
Services wrap this in their init_tracing(config). propagator_factories lets a
service add datadog/sw8 style propagators; span_processors (extra processors,
e.g. a second exporter) are added after the batch processor; sampler replaces the one
built from config.sampler/sampler_ratio (remote or vendor samplers). With
config.set_global the provider and propagator become the
OpenTelemetry globals (the API allows one global provider per process: is_global on
the handle tells whether it took effect).
inject ¶
inject(carrier: MutableMapping[str, str] | None = None, *, context: Context | None = None, propagator: TextMapPropagator | None = None) -> MutableMapping[str, str]
Write the current (or given) context into carrier and return it.
Uses propagator or the global propagator (set by init_tracing_with).
inject_headers ¶
inject_headers(headers: MutableMapping[str, str] | None = None, *, context: Context | None = None, propagator: TextMapPropagator | None = None) -> MutableMapping[str, str]
Add propagation headers (traceparent, baggage …) to headers and return them.
instrument_requests_session ¶
instrument_requests_session(session: Session, *, tracer: Tracer | None = None, propagator: TextMapPropagator | None = None) -> requests.Session
Trace every request sent through session (CLIENT span, headers injected).
Wraps session.send once; calling it again on the same session is a no-op.
log_row_to_record ¶
Convert a raw log row ({timestamp, data} from query_range or live tail).
Severity comes from severity_number (fallback: severity_text); attributes merge
attributes_string|number|bool; resource is resources_string; empty trace/span ids
are absent. A structured body is returned as a JSON string.
mark_span_error ¶
Set ERROR status on span and record the exception (when error is one).
micros_to_datetime ¶
Exact UTC datetime for microseconds since the epoch.
normalize_headers ¶
Lower-cased header names → values; accepts mappings and (name, value) pairs incl. bytes.
normalize_span_id ¶
16-char lower-hex span id (left-padded); raises ValidationError when invalid.
normalize_trace_id ¶
32-char lower-hex trace id; 64-bit ids (16 hex chars, ints, 8 bytes) are left-padded.
Raises ValidationError for non-hex, too long or all-zero ids.
otlp_headers ¶
Headers for OTLP exports: config headers, extra and the Cloud ingestion key.
otlp_http_url ¶
Append /v1/<signal> to an OTLP/HTTP base endpoint unless it already ends with it.
parse_bool ¶
Parse true/false/1/0/yes/no/on/off (case-insensitive); raise :class:ValueError.
parse_duration ¶
Parse a duration into seconds.
Accepts integer milliseconds ("1500" → 1.5) or Go-style units ("500ms", "10s",
"2m", "1h", "1m30s"). Raises :class:ValueError for anything else.
parse_retry_after ¶
Parse Retry-After as delta seconds or an HTTP-date; return seconds (>= 0) or None.
parse_timestamp ¶
Parse RFC 3339 (any fraction precision, Z/offset; none = UTC) to aware UTC.
poll_until ¶
poll_until(fn: Callable[[], T | None], *, timeout_s: float = 30.0, interval_s: float = 1.0, predicate: Callable[[T], bool] | None = None, operation: str = 'poll', sleep: Callable[[float], None] = time.sleep, clock: Callable[[], float] = time.monotonic) -> T
Call fn every interval_s until it returns a ready value; return that value.
A value is ready when it is not None and predicate(value) is true (without a
predicate: not an empty container). Retryable SDK errors raised by fn are remembered
and polling continues; other errors propagate. Raises :class:TimeoutError after timeout_s.
quote_filter_string ¶
Quote a string literal: '…' with \ → \\ and ' → \'.
record_exception ¶
Record error as an exception event on the active span (status unchanged).
redact_fields ¶
Recursively redact values whose key looks sensitive in nested mappings and lists.
redact_headers ¶
Copy of headers with sensitive values replaced by [REDACTED].
redact_text ¶
Replace every known secret value (length >= 4) in text with [REDACTED].
redact_url ¶
Redact user info and every query value (all values, not only sensitive ones).
Query names are kept so the URL stays useful in logs.
resolve_window ¶
resolve_window(start: datetime | None, end: datetime | None, now: datetime | None = None) -> tuple[int, int]
(start_ms, end_ms); defaults: end = now, start = one hour before end.
Raises :class:ValidationError when start is not before end.
retry_after_from_headers ¶
retry_after_from_headers(headers: Headers | Mapping[str, str], names: Sequence[str] = RETRY_AFTER_HEADERS, *, now: Callable[[], float] = time.time) -> float | None
Server requested delay in seconds from the first of names that parses.
Retry-After is delta seconds or an HTTP-date. Other headers (X-RateLimit-Reset,
RateLimit-Reset …) may be delta seconds or a Unix timestamp in seconds or milliseconds;
timestamps are turned into the remaining delay.
rfc3339_to_nanos ¶
Parse an RFC 3339 timestamp with up to nanosecond precision into epoch nanoseconds.
Raises :class:ValidationError when invalid.
set_error ¶
Mark the active span as failed: ERROR status, plus an exception event for exceptions.
signoz_tracing_config ¶
signoz_tracing_config(config: TracingConfig | str, signoz: SignozConfig | None = None, **overrides: Any) -> TracingConfig
Merge SigNoz defaults under a :class:TracingConfig.
config may be a TracingConfig or just the service name. Without an explicit
endpoint, the protocol is signoz.otlp_protocol unless the tracing config picks
grpc (http/protobuf is the TracingConfig default and cannot be told apart from
"unset", so it follows the SigNoz config — same rule as Rust and Java). The endpoint defaults
to the configured collector endpoint for the protocol (4318 ↔ 4317 swapped when the
protocols differ), headers get the signoz-ingestion-key (explicit headers win) and TLS
defaults to signoz.tls. signoz defaults to :meth:SignozConfig.from_env.
sort_spans ¶
Spans sorted by start time, then span id.
span_from_readable ¶
Convert a finished OpenTelemetry SDK span (e.g. from an in-memory exporter).
span_kind ¶
SpanKind from an enum member or a name ("server"); None → INTERNAL.
span_row_to_span ¶
Convert a raw span row (requestType: raw, signal: traces); None without ids.
The parent reference SigNoz mirrors into links is dropped; SigNoz does not store link
attributes, so they are always {}.
structlog_processor ¶
Return a structlog processor that sends every event to client.
The event dict passes through unchanged, so put this processor before the renderer.
event becomes the body, level (or the method name) the severity, timestamp
(TimeStamper output, epoch float or datetime) the timestamp; exc_info and a rendered
exception become exception.* attributes; every other key is an attribute.
to_attr_value ¶
Convert any Python value to an attribute value (unknown types become str(value)).
to_epoch_ms ¶
Epoch milliseconds of a datetime (naive values are UTC).
to_seconds ¶
Normalise a duration given as float seconds or :class:datetime.timedelta.
trace_fields ¶
{"trace_id": …, "span_id": …} of the active span for log fields; {} without one.
traced ¶
traced(name_or_fn: str | Callable[P, R] | None = None, /, *, attributes: Attributes = None, kind: SpanKindLike = None, tracer: Tracer | None = None) -> Callable[P, R] | Callable[[Callable[P, R]], Callable[P, R]]
Decorator that runs a function (sync or async def) inside a span.
@traced, @traced() and @traced("name") all work; the default span name is the
function's qualified name. Async generators are not supported (the span would end before
iteration).
truncate ¶
Truncate text to limit characters (adds an ellipsis when cut).
user_agent ¶
grow-codes-<key>/<version> (python) plus <suffix> when given.
wait_for_log ¶
wait_for_log(client: LogQueryClient, query: str, *, timeout_s: float | timedelta = 30.0, interval_s: float | timedelta = 1.0, predicate: RecordPredicate | None = None) -> LogRecord
Poll client.search(query) until a record (matching predicate) shows up.
wait_for_trace ¶
wait_for_trace(client: TraceQueryClient, trace_id: str, *, timeout_s: float | timedelta = 30.0, interval_s: float | timedelta = 1.0, min_spans: int = 1, normalize: Callable[[str], str] | None = normalize_trace_id) -> Trace
Poll get_trace until the trace has min_spans spans; TimeoutError otherwise.
NotFoundError and retryable errors count as "not there yet". normalize turns
trace_id into the id passed to get_trace (default: 32-char lower hex); None
passes native ids verbatim.
waterfall_span_to_span ¶
Convert one waterfall span; None without ids.
time_unix is in milliseconds, so start_micros has millisecond precision. Links come
from references with refType: FOLLOWS_FROM.
waterfall_to_trace ¶
Convert the data of /api/v4/traces/{id}/waterfall.
with_span ¶
with_span(name: str, fn: Callable[[Span], R] | None = None, *, attributes: Attributes = None, kind: SpanKindLike = None, links: Sequence[Link | SpanContext] | None = None, tracer: Tracer | None = None) -> R | AbstractContextManager[otel_trace.Span]
Run fn(span) inside a new current span, or return a context manager without fn.
The span always ends. An exception is recorded, sets ERROR status and is re-raised
(cancellation, asyncio.CancelledError or the SDK CancelledError, is not). The
context-manager form also works inside async code (context is kept in contextvars).