Repository navigation
Expand file tree
/
Copy path.env.example
More file actions
90 lines (76 loc) · 3.4 KB
/
Copy path.env.example
File metadata and controls
90 lines (76 loc) · 3.4 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
# .env.example — WP04 T035
#
# Copy this file to `.env` and fill in the placeholders. The
# docker-compose.yml mounts `.env` into both `api` and
# `ingestion` containers via `env_file`.
#
# NEVER commit a real `.env` -- only this template. See
# AGENTS.md §7.2 (Secrets handling) and §7.1 (SecretScrubber).
#
# Mandatory for any real answer generation:
OPENAI_API_KEY=sk-REPLACE_ME
# Source page to ingest (the ingestion Job fetches and chunks
# this URL into Chroma). The URL is single-valued because the
# ingestion pipeline is single-source by design (WP03).
SOURCE_URL=https://example.com/internet
# LLM provider for the answer step. `openai` is the default
# shipped in WP02 + WP04; `fake` is for local development
# without a key (the application layer short-circuits to
# `tests.fakes.answering.answer_generator.FakeAnswerGenerator`).
# See plan.md § Phase 0.5 for the create_agent opt-in path
# (future WP).
ANSWERER_BACKEND=openai
# Embedder provider. `local` uses sentence-transformers
# (all-MiniLM-L6-v2, 384 dims, no API key, default for
# ingestion). `openai` uses text-embedding-3-small (1536 dims,
# needs OPENAI_API_KEY above). Both adapters are wired in
# WP03 + WP04.
EMBEDDER_BACKEND=local
# Vector store. In docker-compose the host is `chroma` (the
# service name); the port matches the chroma container. For a
# local non-Docker run, set CHROMA_HOST=localhost.
CHROMA_HOST=chroma
CHROMA_PORT=8000
# Optional Chroma auth token (sent as an Authorization
# header by ChromaVectorStore). Leave empty in local dev.
CHROMA_AUTH_TOKEN=
# Optional embedder API key (only used when EMBEDDER_BACKEND=openai).
# Embedding_API_KEY=
# Confidence threshold for ThresholdLowConfidencePolicy (cosine
# similarity, [0.0, 1.0]). Below this the agent refuses with
# the spec-mandated message (FR-009).
LOW_CONFIDENCE_THRESHOLD=0.5
# WP06 — semantic page analyzer (LLM-driven) + chunker backend
# selector. ``openai`` enables OpenAIPageAnalyzer (uses
# OPENAI_PAGE_ANALYZER_MODEL below); ``none`` skips the analyze
# step entirely (the chunker receives no PageStructure and falls
# back to the WP03 fixed-size path).
ANALYZER_BACKEND=openai
# WP06 — chunker backend. ``hybrid`` consumes PageStructure
# (one chunk per FAQ + semantic chunking for the rest); set
# ``fixed_size`` to revert to the WP03 FixedSizeChunker.
CHUNKER_BACKEND=hybrid
# WP06 — OpenAI chat-completions model used by the page
# analyzer. Defaults to gpt-4o-mini for cost; gpt-4o yields
# higher-quality JSON at higher latency/cost.
OPENAI_PAGE_ANALYZER_MODEL=gpt-4o-mini
# Lock directory for the ingestion Job (TTL-keyed file lock,
# see AGENTS.md §4.5). On Linux the host directory must be
# writable by the container UID (uid 1001).
LOCK_DIR=/var/run/support-bot
# Request ID for the ingestion run (default = random uuid4 if
# unset). The lock filename is `<LOCK_DIR>/<REQUEST_ID>.lock`,
# so the same ID is in logs, spans, and on-disk artefacts.
# `RUN_ID` is a deprecated alias (AGENTS.md §6.3).
REQUEST_ID=
# Logging level (structlog: DEBUG / INFO / WARNING / ERROR).
LOG_LEVEL=INFO
# OpenTelemetry: leave OTEL_EXPORTER_OTLP_ENDPOINT empty to
# disable the exporter (spans are still created and attached
# to logs but never sent). For dev/staging point at a local
# collector, e.g. http://otel-collector:4318.
OTEL_EXPORTER_OTLP_ENDPOINT=
OTEL_SERVICE_NAMESPACE=support-bot
OTEL_DEPLOYMENT_ENVIRONMENT=local
OTEL_TRACES_SAMPLER=parentbased_traceidratio
OTEL_TRACES_SAMPLER_ARG=1.0