Struct Config 

pub struct Config {

    pub port: u16,
    pub worker_token: String,
    pub github_webhook_secret: String,
    pub bitbucket_webhook_secret: String,
    pub gitlab_webhook_secret: String,
    pub openrouter_api_key: String,
    pub openrouter_model: String,
    pub openrouter_model_explore: String,
    pub openrouter_base_url: String,
    pub openrouter_max_tokens: u32,
    pub openrouter_temperature: f32,
    pub openrouter_timeout_secs: u64,
    pub openrouter_max_retries: u32,
    pub max_diff_chars: usize,
    pub include_globs: Vec<String>,
    pub exclude_globs: Vec<String>,
    pub vendored_globs: Vec<String>,
    pub file_bundling: bool,
    pub self_critique: bool,
    pub min_confidence: u8,
    pub max_findings: usize,
    pub reanchor_findings: bool,
    pub agentic: bool,
    pub max_turns: usize,
    pub max_history_chars: usize,
    pub review_on_update: bool,
    pub github_token: String,
    pub github_api_base: String,
    pub bitbucket_email: String,
    pub bitbucket_token: String,
    pub bitbucket_api_base: String,
    pub gitlab_token: String,
    pub gitlab_api_base: String,
    pub comment_marker: String,
    pub user_agent: String,
    pub http_referer: String,
    pub x_title: String,
    pub extra_system_prompt: String,
    pub structural_context: bool,
    pub structural_max_files: usize,
    pub complexity_metrics: bool,
    pub complexity_min_cyclomatic: u32,
    pub blast_radius: bool,
    pub blast_max_symbols: usize,
    pub blast_max_refs: usize,
    pub ci_status: bool,
    pub cve_scan: bool,
    pub cve_max_packages: usize,
    pub osv_api_base: String,
}

Resolved runtime configuration.

Fields

port: u16

worker_token: String

github_webhook_secret: String

Shared secret GitHub signs webhook deliveries with (X-Hub-Signature-256).

bitbucket_webhook_secret: String

Shared secret Bitbucket signs webhook deliveries with (X-Hub-Signature).

gitlab_webhook_secret: String

Shared token GitLab sends webhook deliveries with (X-Gitlab-Token). Unlike GitHub/Bitbucket this is a plain token compared verbatim, not an HMAC.

openrouter_api_key: String

OpenRouter (or any OpenAI-compatible) API key. Resolved from OPENROUTER_API_KEY, falling back to LLM_API_KEY — the alias lets Ollama/vLLM/local OpenAI-compatible servers reuse a generic var name.

openrouter_model: String

Synthesis model: writes the final review findings (quality matters here).

openrouter_model_explore: String

Exploration model: drives the agentic tool loop (grep/read_file/list_dir) to gather context. Cheaper — it navigates files, it doesn’t judge.

openrouter_base_url: String

Base URL of the OpenAI-compatible chat-completions API. Resolved from LLM_BASE_URL, then OPENROUTER_BASE_URL, defaulting to OpenRouter — the LLM_BASE_URL alias lets Ollama/vLLM/local OpenAI-compatible servers work.

openrouter_max_tokens: u32

openrouter_temperature: f32

openrouter_timeout_secs: u64

Whole-request timeout for a single model call. The hand-rolled loop had none, so a stalled provider hung the entire review.

openrouter_max_retries: u32

Retries for a transient failure (429 / 5xx) on one call. The hand-rolled loop had none, so a single 429 discarded the whole review.

max_diff_chars: usize

include_globs: Vec<String>

Glob patterns of files to INCLUDE in the diff before it’s sent to the LLM. Empty means include everything (subject to exclude_globs).

exclude_globs: Vec<String>

Glob patterns of files to EXCLUDE from the diff (lockfiles, generated, vendored, minified) — drops noise and saves tokens before the LLM call.

vendored_globs: Vec<String>

Glob patterns marking vendored third-party source. Diff-hygiene (class D) findings are suppressed inside these paths — bulk-committing upstream code is the intent of a vendoring PR, not a defect. Empty means the conventional defaults (crate::diff::DEFAULT_VENDORED_DIRS).

file_bundling: bool

When packing a large diff to the size budget, group related changed files (a source + its test, i18n siblings) into one unit so they stay adjacent and pack together instead of being scattered by priority. Fail-open.

self_critique: bool

Run a second, skeptical “self-critique” pass that removes false positives and out-of-scope nits from the findings before posting.

min_confidence: u8

Drop findings whose model-reported confidence is below this threshold.

max_findings: usize

Hard cap on the number of findings posted (after sorting by severity).

reanchor_findings: bool

Re-anchor a finding whose line just missed a real diff line: snap it to the nearest diff line within a small window when that line’s code matches what the finding references, so a small drift posts inline instead of folding into the summary. Conservative + fail-open.

agentic: bool

Use the agentic reviewer: clone the repo and let the model investigate cross-file context with tools, instead of a single diff-only call.

max_turns: usize

Max tool-loop turns for the agentic reviewer (cost guard).

max_history_chars: usize

Char budget for accumulated tool results in the agent conversation; older results are elided once newer ones fill it (bounds per-turn tokens).

review_on_update: bool

Re-review when new commits are pushed to an open PR (GitHub synchronize / Bitbucket pullrequest:updated). Off by default so iterating on a PR doesn’t trigger a fresh (expensive) review per push — opened/reopened/ ready_for_review always review; use the manual POST /review endpoint to re-review on demand.

github_token: String

github_api_base: String

bitbucket_email: String

bitbucket_token: String

bitbucket_api_base: String

gitlab_token: String

gitlab_api_base: String

comment_marker: String

Signature appended to every comment the bot posts and used as the dedupe key to find/update its own comments. Injected so the library carries no hardcoded bot identity.

user_agent: String

User-Agent header sent on provider (GitHub) API requests.

http_referer: String

HTTP-Referer header sent to OpenRouter (attribution).

x_title: String

X-Title header sent to OpenRouter (attribution).

extra_system_prompt: String

Extra system-prompt text appended to the built-in prompts. Lets a consumer inject a large conventions block (e.g. via a file baked into its image) without changing the library.

structural_context: bool

Compute “structural context” (the enclosing function/symbol of each changed line) and inject it into the review prompt so the model knows each change’s scope. Fully fail-open — never blocks a review.

structural_max_files: usize

Max number of files to fetch + parse for structural context (cost guard).

complexity_metrics: bool

Surface deterministic complexity metrics (cyclomatic + cognitive, A–F grade) for the functions a change touches, computed with tree-sitter alongside the structural context. A cheap, LLM-free risk signal. Fully fail-open.

complexity_min_cyclomatic: u32

Only report a changed function when its cyclomatic complexity is at least this — keeps the block a high-signal risk flag, not noise for trivial code.

blast_radius: bool

Compute a “blast radius” for the agentic reviewer: from the clone, find the callers and tests of each changed symbol and seed the prompt with them (also exposes a references tool). Agentic path only — needs the clone. Fail-open.

blast_max_symbols: usize

Max number of changed symbols to expand into a blast radius (cost guard).

blast_max_refs: usize

Max call sites listed per symbol per bucket (callers / tests) (cost guard).

ci_status: bool

Fetch the head commit’s CI results (GitHub check runs / Bitbucket build statuses) with the PR metadata and surface them in the prompt, so the reviewer can’t assert a broken build that CI already decided.

One extra API call per review. Off is for tokens near their rate limit; with it off the reviewer simply sees no CI block. Fully fail-open either way.

cve_scan: bool

Scan changed lockfiles for known-vulnerable dependencies via OSV.dev and append advisories to the review summary. Fully fail-open — never blocks a review.

cve_max_packages: usize

Max distinct packages queried against OSV per review (cost/fan-out guard).

osv_api_base: String

Base URL of the OSV.dev API (override for a mirror or a test double).

Implementations

impl Config

pub fn from_env() -> Self

Build the config from environment variables, applying defaults.

pub fn with_repo_overrides(&self, rc: &RepoConfig) -> Config

Return a clone of this config with any fields set in rc overridden.

Only Some(..) fields of the per-repo [RepoConfig] take effect; the rest keep the env-derived value. rc.instructions is appended to Config::extra_system_prompt (newline-separated) rather than replacing it, so a consumer’s baked-in conventions block and the repo’s own instructions both reach the model.

Examples

let base = Config::from_env();
let rc = RepoConfig { min_confidence: Some(80), ..Default::default() };
let effective = base.with_repo_overrides(&rc);
assert_eq!(effective.min_confidence, 80);

Trait Implementations

impl Clone for Config

fn clone(&self) -> Config

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.