aria/conf_template.yaml at main · aria-aiops/aria · GitHub

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
# ──────────────────────────────────────────────────────────────────────────────
# ARIA — Configuration template
# Copy this file to conf.yaml and fill in your values. Never commit conf.yaml.
#
# conf.yaml holds non-secret runtime configuration. Actual secrets (passwords,
# API keys, tokens, SSH keys) belong in .env / your secrets manager.
#
# NOTE: The codebase currently reads configuration from environment variables.
# conf.yaml defines the target configuration model. Code integration (replacing
# env var reads with conf.yaml reads) is a pre-S6 task.
# ──────────────────────────────────────────────────────────────────────────────

runtime:
  operating_mode: inform    # inform | hitm | autonomous — only 'inform' is implemented
                            # in Phase 1.5; the others raise NotImplementedError until
                            # their phase ships (hitm: Phase 2, autonomous: Phase 3)
  vault_backend: env        # env | gcp | hashicorp | aws | azure
                            # 'env' reads secrets from environment variables (local dev)
                            # 'gcp' uses GCP Secret Manager via ADC (container deployments)
  log_format: human         # human | json — human is coloured structlog for terminals;
                            # json is machine-readable for log aggregators (Splunk, ELK)
  log_dir: logs/            # directory for rolling log files (overridden by ARIA_LOG_DIR)

runs:
  db_path: data/runs.db     # SQLite file for run history (monitoring API + dashboard)

servicenow:
  instance: <your-instance>.service-now.com   # e.g. dev12345.service-now.com
  user: <your-service-account>                # ServiceNow service account username
  assignment_group: <your-assignment-group>   # e.g. "Data Platform OPS"
  cmdb_rel_type: "Members::Member of"         # override if your instance uses a different type

llm:
  provider: anthropic       # anthropic | claude_code | vertex_ai
                            # anthropic:   direct Anthropic API — needs ANTHROPIC_API_KEY
                            # claude_code: local Claude Code CLI — local dev only (security risk in prod, #84)
                            # vertex_ai:   GCP Vertex AI via ADC — use for container deployments
  mode: modular             # modular | global
  global_model: ""          # used only when mode=global, e.g. claude-sonnet-4-6
  agents:
    agent1: <model-id>      # e.g. claude-sonnet-4-6
    agent2: ""              # optional — enables LLM query planning in Agent 2
    agent3: <model-id>
    agent4: ""              # optional — unused in Phase 1; activates Phase 2 response interpretation

connectors:
  log: memory               # memory | gcp
  queue: memory             # memory (only option in Phase 1)
  state_store: memory       # memory (only option in Phase 1)

gcp:                        # only required when connectors.log = gcp
  project_id: <your-gcp-project-id>
  region: <your-gcp-region>
  gcs_bucket_logs: <your-gcs-bucket-name>
  bq_log_dataset: <your-bigquery-dataset>

cdp:
  ssh_user: hadoop           # OS user with read access to log directories below
  ssh_key_secret: CDP_SSH_KEY  # vault key name for the SSH private key PEM.
                               # EnvVarVault: reads os.environ[ssh_key_secret] directly.
                               # GCPSecretManagerVault: normalises underscores→hyphens and
                               # prepends 'aria-', so 'CDP_SSH_KEY' → GCP secret 'aria-cdp-ssh-key'.
                               # UC1 (TF): TF provisions secret 'aria-uc1-ssh-private-key'.
                               # Set ssh_key_secret: CDP_UC1_SSH_PRIVATE_KEY to match, or
                               # rename the TF secret to 'aria-cdp-ssh-key'.
  log_dirs:                  # directories searched for logs on CDP cluster nodes.
                             # UC1 (TF Hadoop VMs) use subdirectory paths — set these
                             # at deployment time from TF startup script:
    - /var/log/hadoop/hdfs   # TF UC1 NameNode / DataNode logs
    - /var/log/hadoop/yarn   # TF UC1 ResourceManager / NodeManager logs
    - /var/log/hive
    - /var/log/spark
    - /var/log/kafka
    - /var/log/zookeeper
    - /var/log/oozie
    - /var/log/nifi

knowledge_base:
  analyser_kb_dir: data/knowledge_base/analyser_kb  # labeled log excerpts for Agent 3 few-shot prompting
                                                     # doubles as fine-tuning corpus for the future
                                                     # specialist Agent 3 model (roadmap: post-POC)

slack:
  channel_id: <your-slack-channel-id>   # channel where Agent 4 posts notifications