> ## Documentation Index
> Fetch the complete documentation index at: https://praison.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Config from Environment
> Inject the full CLI config from an environment variable — no file on disk
Supply the entire user-config layer from an environment variable, so containers and CI runners need nothing written to disk.
```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
graph LR
subgraph "Config Resolution"
A[🧩 PRAISONAI_CONFIG_CONTENT] --> R{🔍 Resolver}
B[📄 PRAISONAI_CONFIG] --> R
C[🏠 Global file] --> R
D[📁 Project file] --> R
E[⚙️ Defaults] --> R
R --> O[✅ Resolved Config]
end
classDef inline fill:#8B0000,stroke:#7C90A0,color:#fff
classDef path fill:#189AB4,stroke:#7C90A0,color:#fff
classDef file fill:#F59E0B,stroke:#7C90A0,color:#fff
classDef default fill:#6366F1,stroke:#7C90A0,color:#fff
classDef result fill:#10B981,stroke:#7C90A0,color:#fff
class A inline
class B path
class C,D file
class E default
class O result
```
## Quick Start
Pass the whole config as an inline blob — ideal for Docker and CI where no file exists:
```bash theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
export PRAISONAI_CONFIG_CONTENT='
agent:
model: gpt-4o
output: verbose
'
praisonai "Summarise today's changes"
```
Point at a specific config file instead of the discovered global/project files:
```bash theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
export PRAISONAI_CONFIG=/etc/praisonai/config.yaml
praisonai "Summarise today's changes"
```
Values support `${VAR}`, `{env:VAR}`, and `{file:...}` so secrets stay out of the blob:
```bash theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
export OPENAI_API_KEY=sk-...
export PRAISONAI_CONFIG_CONTENT='
agent:
model: gpt-4o
api_key: ${OPENAI_API_KEY}
system_prompt: "{file:/run/secrets/prompt.txt}"
'
praisonai "Run the pipeline"
```
***
## How It Works
Both JSON and YAML parse through `yaml.safe_load`, then interpolation runs before the blob joins the resolution ladder.
```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
sequenceDiagram
participant Env
participant Resolver
participant Interpolate
Env->>Resolver: PRAISONAI_CONFIG_CONTENT / PRAISONAI_CONFIG
Resolver->>Resolver: yaml.safe_load (JSON or YAML)
Resolver->>Interpolate: ${VAR} / {env:VAR} / {file:...}
Interpolate-->>Resolver: resolved values
Resolver-->>Env: merged into user-config layer
```
Both variables fill the **user-config layer** — the same slot the discovered global/project file would occupy. When either is set, no global or project file is read for that layer.
| Setting | Purpose |
| -------------------------- | -------------------------------------- |
| `PRAISONAI_CONFIG_CONTENT` | Inline JSON/YAML blob parsed in memory |
| `PRAISONAI_CONFIG` | Explicit path to a config file |
***
## Precedence
Higher layers win. The env-config sources sit in the user-config layer, below per-key env vars and CLI flags.
| Layer | Source | Wins Over |
| ----- | ---------------------------------------------------- | ----------------- |
| 1 | Defaults | — |
| 2 | Managed defaults | Defaults |
| 3 | **User-config layer** (env-config or file discovery) | Managed defaults |
| 4 | Per-key environment variables | User-config layer |
| 5 | CLI arguments | Per-key env vars |
| 6 | Managed policy | Everything |
Within the user-config layer, inline content beats an explicit path, which beats global-file discovery, which beats project-file discovery:
```
PRAISONAI_CONFIG_CONTENT > PRAISONAI_CONFIG > global file > project file
```
When the blob is invalid JSON/YAML, or `PRAISONAI_CONFIG` points at a missing file, the resolver warns and falls back to normal file discovery. Behaviour is unchanged when neither variable is set.
***
## Common Patterns
### Docker / Kubernetes secret mount
Mount the config as a secret and export it — nothing is written to the image:
```yaml theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
env:
- name: PRAISONAI_CONFIG_CONTENT
valueFrom:
secretKeyRef:
name: praisonai-config
key: config.yaml
```
### GitHub Actions runner
Ship one variable to the job; no `~/.praisonai/` needs to exist:
```yaml theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
- name: Run agent
env:
PRAISONAI_CONFIG_CONTENT: |
agent:
model: gpt-4o
output: silent
run: praisonai "Analyse the diff"
```
### Ephemeral container with no writable home
Inline content avoids any filesystem write on a read-only container:
```bash theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
docker run --read-only \
-e PRAISONAI_CONFIG_CONTENT="$(cat config.yaml)" \
praisonai "Run once and exit"
```
***
## User Interaction Flow
A team lead ships one environment variable to the CI container. The runner has no `~/.praisonai/` and never writes one — the resolver reads `PRAISONAI_CONFIG_CONTENT`, interpolates any `${VAR}` secrets from the job's environment, and runs the agent with the team's model and output settings. Removing the variable restores default file discovery with no code change.
`PRAISONAI_AUTH_CONTENT` is the credentials-only sibling. Use it to inject the auth layer the same way, so a single CI secret set can supply both config and credentials with nothing on disk.
***
## Best Practices
Reference secrets with `${VAR}` or `{file:/run/secrets/...}` instead of pasting them inline, so the config blob stays safe to log.
`PRAISONAI_CONFIG_CONTENT` needs no writable filesystem, making it ideal for read-only containers and ephemeral CI.
When a config file is already mounted, `PRAISONAI_CONFIG` points at it directly and skips global/project discovery.
An invalid blob or missing path warns and falls back to file discovery, so a typo never hard-fails a run.
***
## Related
Control agent output verbosity and formatting
Where config, data, state, and cache live