tools-ask-ai-core — IMP FORGE

Shared client for the multi-AI bridge (ask-chatgpt / ask-gemini / ask-ais)

Tier	tools
Role	unclassified (baselined)
Path	`crates/tools/ask-ai-core`
Edition	2021
Targets	tools_ask_ai_core
Public items	43 across 1 module
Tests	31

What it is for

Shared client for the multi-AI bridge.

Exposes:

AiProvider — the supported AI backends.
AskRequest / AskResponse — the request/response shape.
ask() — synchronous single-provider call against the bridge HTTP

endpoint at /api/ask-{chatgpt,gemini}.

ask_many() — parallel fan-out to multiple providers via OS threads

(no async runtime required). Returns per-provider AskResponse entries; the caller interprets success/failure per the Gate 1.5 Round 1 Q4 decision (tools-ask-ais exit code semantics).

Capabilities

`verb:ask`

Shared client for the multi-AI bridge.

Item
`fn ask(api_url : & str, api_key : & str, req : & AskRequest) -> AskResponse`
`fn ask_with_recovery(api_url : & str, api_key : & str, req : & AskRequest) -> AskResponse`
`fn ask_many(api_url : & str, api_key : & str, providers : & AiProvider, req : & AskRequest,) -> Vec <AskResponse>`

`verb:busy`

Shared client for the multi-AI bridge.

Item
`fn busy_retry_after(body : & serde_json::Value, max_wait_secs : u64) -> Option <Duration>`

`verb:constants`

Shared client for the multi-AI bridge.

Item
`pub const DEFAULT_API_URL: & str`
`pub const DEFAULT_API_KEY: & str`
`pub const DEFAULT_TIMEOUT_SECS: u64`
`pub const COLLECT_TIMEOUT_SECS: u64`
`pub const DEFAULT_MAX_BUSY_WAIT_SECS: u64`
`pub const BUSY_POLL_CEILING_SECS: u64`

`verb:default`

Shared client for the multi-AI bridge.

Item
`fn default_api_url() -> String`
`fn default_api_key() -> String`

`verb:encode`

Shared client for the multi-AI bridge.

Item
`fn encode_image(path : & str) -> Result <ImageAttachment, AskError>`

`verb:max`

Shared client for the multi-AI bridge.

Item
`fn max_busy_wait_secs() -> u64`

`verb:session`

Shared client for the multi-AI bridge.

Item
`fn session_id() -> String`

`AiProvider`

Shared client for the multi-AI bridge.

Item
`pub enum AiProvider`
`AiProvider :: fn as_str(self) -> & 'static str`
`AiProvider :: fn command_type(self) -> & 'static str`
`AiProvider :: fn endpoint_path(self) -> & 'static str`
`AiProvider :: fn display_label(self) -> & 'static str`
`AiProvider :: fn parse(s : & str) -> Option <Self>`

`AskError`

Shared client for the multi-AI bridge.

Item
`pub enum AskError`

`AskRequest`

Shared client for the multi-AI bridge.

Item
`pub struct AskRequest`

`AskResponse`

Shared client for the multi-AI bridge.

Item
`pub struct AskResponse`
`AskResponse :: fn ok(provider : AiProvider, text : String, duration_ms : u128) -> Self`
`AskResponse :: fn error(provider : AiProvider, status : AskStatus, error : String, duration_ms : u128,) -> Self`
`fn format_single(resp : & AskResponse) -> Result <String, String>`
`fn format_labeled(responses : & AskResponse) -> String`
`fn format_json(responses : & AskResponse) -> String`
`fn exit_code_for(responses : & AskResponse) -> i32`

`AskStatus`

Shared client for the multi-AI bridge.

Item
`pub enum AskStatus`
`AskStatus :: fn as_str(self) -> & 'static str`
`AskStatus :: fn is_success(self) -> bool`
`fn is_collectible(status : AskStatus, error : Option <& str>) -> bool`

`ImageAttachment`

Shared client for the multi-AI bridge.

Item
`pub struct ImageAttachment`

`precheck (other)`

Advisory client-side DLP pre-check (Sprint 3.6 T9).

Item
`pub const POLICY_ENV: & str`
`fn load_policy_from_path(path : Option <& str>) -> Result <ScreenPolicy, String>`
`fn load_policy() -> Result <ScreenPolicy, String>`
`fn resolve_prompt_arg(message : & str) -> Result <String, String>`

`CliGateDecision`

Advisory client-side DLP pre-check (Sprint 3.6 T9).

Item
`pub enum CliGateDecision`

`Precheck`

Advisory client-side DLP pre-check (Sprint 3.6 T9).

Item
`pub enum Precheck`
`fn precheck_prompt(prompt : & str, policy : & ScreenPolicy) -> Precheck`
`fn cli_precheck(prompt : & str, force_send : bool) -> CliGateDecision`

How to use it

No examples/ target and no doctest in this crate's rustdoc. The tests listed under Verification are the closest executable usage.

Module structure

tools_ask_ai_core

precheck

Public surface

`crate root`

Item	What it is
`pub const DEFAULT_API_URL: & str`	—
`pub const DEFAULT_API_KEY: & str`	—
`pub const DEFAULT_TIMEOUT_SECS: u64`	300s: consensus-scale prompts on slow provider modes routinely outlast the old 180s (three timeouts on 2026-08-08 alone)
`pub const COLLECT_TIMEOUT_SECS: u64`	Budget for the post-timeout collect retry (extension-side poll).
`pub const DEFAULT_MAX_BUSY_WAIT_SECS: u64`	How long to keep waiting for a provider tab another session holds, before giving up
`pub const BUSY_POLL_CEILING_SECS: u64`	Longest single sleep between retries while a provider tab is held
`pub enum AiProvider`	—
`AiProvider :: fn as_str(self) -> & 'static str`	—
`AiProvider :: fn command_type(self) -> & 'static str`	The extension-side command type string.
`AiProvider :: fn endpoint_path(self) -> & 'static str`	The bridge HTTP endpoint path (sans scheme/host).
`AiProvider :: fn display_label(self) -> & 'static str`	Display label used in multi-provider output (`=== CHATGPT ===`).
`AiProvider :: fn parse(s : & str) -> Option <Self>`	—
`pub struct AskRequest`	—
`pub struct ImageAttachment`	—
`pub enum AskStatus`	Status of a single-provider ask.
`AskStatus :: fn as_str(self) -> & 'static str`	—
`AskStatus :: fn is_success(self) -> bool`	—
`pub struct AskResponse`	—
`AskResponse :: fn ok(provider : AiProvider, text : String, duration_ms : u128) -> Self`	—
`AskResponse :: fn error(provider : AiProvider, status : AskStatus, error : String, duration_ms : u128,) -> Self`	—
`pub enum AskError`	—
`fn encode_image(path : & str) -> Result <ImageAttachment, AskError>`	Load and base64-encode an image, returning an `ImageAttachment` ready to attach to an `AskRequest`
`fn default_api_url() -> String`	Read the bridge URL from `ASK_CHATGPT_API` env var, falling back to the default localhost bridge.
`fn default_api_key() -> String`	Read the bridge API key from `ASK_CHATGPT_KEY` env var, falling back to the dev default.
`fn session_id() -> String`	Synchronously ask a single AI provider
`fn busy_retry_after(body : & serde_json::Value, max_wait_secs : u64) -> Option <Duration>`	How long to wait before retrying, given a 409 body — or `None` if this is not a provider-busy refusal and so not something waiting can fix
`fn max_busy_wait_secs() -> u64`	The ceiling on total time spent waiting out other sessions.
`fn ask(api_url : & str, api_key : & str, req : & AskRequest) -> AskResponse`	—
`fn is_collectible(status : AskStatus, error : Option <& str>) -> bool`	True when a failed ask is worth a collect retry: the send reached the provider and generation likely FINISHED after our clock ran out — a timeout, a gateway 504, or a bridge-side timeout message
`fn ask_with_recovery(api_url : & str, api_key : & str, req : & AskRequest) -> AskResponse`	`ask` plus the timeout-recovery pass: when the ask fails collectibly (see `is_collectible`), one `collect_response` attempt recovers the answer stranded in the provider tab
`fn ask_many(api_url : & str, api_key : & str, providers : & AiProvider, req : & AskRequest,) -> Vec <AskResponse>`	Fan out to multiple providers in parallel (one OS thread per provider)
`fn format_single(resp : & AskResponse) -> Result <String, String>`	Format a single-provider response as plain text (for `ask-chatgpt` / `tools-ask-gemini`)
`fn format_labeled(responses : & AskResponse) -> String`	Format multi-provider responses as labeled blocks (for `tools-ask-ais` default mode)
`fn format_json(responses : & AskResponse) -> String`	Format multi-provider responses as a single JSON object keyed by provider name (for `tools-ask-ais --json`)
`fn exit_code_for(responses : & AskResponse) -> i32`	Pick an appropriate process exit code for `tools-ask-ais` responses per Gate 1.5 Round 1 Q4: total success or partial success → 0, total failure → 1.

`precheck`

Item	What it is
`pub const POLICY_ENV: & str`	Env var naming the local policy TOML for the client pre-check (same file format the server loads; see the platform-privacy-scan-api README).
`pub enum Precheck`	Outcome of the local advisory pre-check.
`fn load_policy_from_path(path : Option <& str>) -> Result <ScreenPolicy, String>`	Load the pre-check policy from an explicit path, or the built-in default policy when `path` is `None`
`fn load_policy() -> Result <ScreenPolicy, String>`	Load the pre-check policy from the `DLP_GATE_POLICY` env var
`fn precheck_prompt(prompt : & str, policy : & ScreenPolicy) -> Precheck`	Run the advisory pre-check on `prompt` against `policy`
`pub enum CliGateDecision`	The full CLI-side decision: load the local policy (env), pre-check, and fold in `--force-send`
`fn cli_precheck(prompt : & str, force_send : bool) -> CliGateDecision`	Apply the advisory pre-check for a CLI invocation
`fn resolve_prompt_arg(message : & str) -> Result <String, String>`	Resolve a CLI prompt argument: `-` reads the prompt from STDIN

Re-exports. Exported here, defined elsewhere.

Export	Defined in
`ScreenPolicy`	`infrastructure_dlp_detect::ScreenPolicy`

Boundary

Reaches into infrastructure.

Shares tier tools with 84 other crates: tools-advisory-reach, tools-archive-guard, tools-artifact-scaffold, tools-ask-ais, tools-ask-gemini, tools-book, tools-book-report, tools-boundary-scan, … (84 total).

_What this crate deliberately does NOT own is a judgment. No committed registry records one for it, so none is stated here._

Where it sits

Tier (ontology)	`tools`
Architectural role (taxonomy)	`unclassified (baselined)`
Location	`crates/tools/ask-ai-core`
Vocabulary in force (lexicon)	`current`

Tier flow. Which tiers this crate's own edges cross.

flowchart LR
  n_tools["tools"] --> n_infrastructure["infrastructure"]

Dependencies

Runtime, in this workspace.

Crate	Tier	Optional	Only on
`infrastructure-dlp-detect`	infrastructure	no	always

Runtime, from outside the workspace.

Crate	Requirement	Features	Optional	Only on
`base64`	`^0.22`	—	no	always
`serde`	`^1`	`derive`	no	always
`serde_json`	`^1`	—	no	always
`thiserror`	`^2`	—	no	always
`ureq`	`^2.10`	—	no	always

Development. None.

Build. None.

Depended on by. 2 workspace crates.

Signal flow — what reaches this crate, and what it reaches.

flowchart LR
  n_tools_ask_ais["tools-ask-ais"] -->|uses| SELF
  n_tools_ask_gemini["tools-ask-gemini"] -->|uses| SELF
  SELF["tools-ask-ai-core"]
  SELF -->|runtime| n_infrastructure_dlp_detect["infrastructure-dlp-detect"]
  classDef self fill:#1f883d,stroke:#1f883d,color:#fff;
  class SELF self;

Feature flags

No Cargo features are defined: every capability is unconditional, so no consumer can receive a half-wired crate.

Targets

Kind	Name	Source
lib	`tools_ask_ai_core`	`src/lib.rs`

Error model

Error type	Named by
`AskError`	`encode_image`

Operational characteristics

Property	Evidence
async public surface	none detected
async runtime	none detected
database access	none detected
network I/O	none detected
unsafe code	none detected
environment variables	yes

No unsafe block, unsafe fn, unsafe impl or unsafe trait was found by the parser anywhere in this crate's source.

Configuration

Variable	Read in
`ASK_AI_MAX_WAIT_SECS`	`src/lib.rs`
`ASK_AI_SESSION_ID`	`src/lib.rs`
`ASK_CHATGPT_API`	`src/lib.rs`
`ASK_CHATGPT_KEY`	`src/lib.rs`

2 workspace crates depend on this one: tools-ask-ais, tools-ask-gemini.

Verification

Kind	Count
Unit tests	31
Integration tests	0
Examples	0
Doctests	0

Evidence by module. How often each public module is named by something executable.

Module	Tests	Examples	Consumers
`crate root`	26	0	19
`precheck`	8	0	6

What the tests establish, by name:

a_lease_about_to_lapse_is_not_over_waited — src/lib.rs
a_long_lease_is_polled_rather_than_slept_out — src/lib.rs
a_missing_expiry_still_produces_a_short_wait_rather_than_none — src/lib.rs
a_nonsense_expiry_cannot_park_the_caller — src/lib.rs
a_session_id_is_stable_within_one_process_and_overridable — src/lib.rs
ask_status_success_semantics — src/lib.rs
collectibility_covers_exactly_the_stranded_answer_cases — src/lib.rs
encode_image_rejects_missing_file — src/lib.rs
encode_image_rejects_unsupported_extension — src/lib.rs
exit_code_nonzero_on_total_failure — src/lib.rs
exit_code_zero_on_partial_success — src/lib.rs
exit_code_zero_on_total_success — src/lib.rs
format_json_is_keyed_by_provider — src/lib.rs
format_labeled_includes_both_provider_sections — src/lib.rs
format_labeled_marks_failures_explicitly — src/lib.rs
format_single_error_returns_error_message — src/lib.rs
format_single_success_returns_text — src/lib.rs
only_a_busy_refusal_is_worth_waiting_on — src/lib.rs
provider_metadata_is_consistent — src/lib.rs
provider_parsing — src/lib.rs
auto_redact_policy_is_advisory_client_never_mutates — src/precheck.rs
blocked_prompt_summary_is_masked_only_and_actionable — src/precheck.rs
clean_prompt_passes — src/precheck.rs
cli_precheck_clean_prompt_proceeds — src/precheck.rs
cli_precheck_force_send_defers_to_server — src/precheck.rs
cli_precheck_refuses_blocked_prompt — src/precheck.rs
email_prompt_is_advisory_not_blocking — src/precheck.rs
load_policy_defaults_when_no_path — src/precheck.rs
load_policy_reports_invalid_toml — src/precheck.rs
load_policy_reports_missing_file — src/precheck.rs
_… 1 more_

Documentation coverage

Measure	Documented	Total
Public items with rustdoc	29	43
Public modules with a `//!` block	1	1

pie showData
    title Public items with rustdoc
    "Documented" : 29
    "No rustdoc detected" : 14

Metrics

Metric	Value
Rust source files	2
Source lines	1306
Code lines	978
Public API items	43
Public modules	1
Tests	31
Examples	0
Cargo features	0
Direct runtime dependencies	6
Workspace reverse dependencies	2

pie showData
    title Public API by kind
    "constant" : 7
    "enum" : 5
    "function" : 19
    "method" : 9
    "struct" : 3

pie showData
    title Rust source composition
    "Code" : 978
    "Blank or comment" : 328

Generation

Rendered by tools-corpus corpus readme from repository evidence alone, renderer schema 2, lexicon current. No model, network service or database was consulted. Regenerate with tools-corpus corpus readme --write; verify with --check.

What it is for

Capabilities

verb:ask

verb:busy

verb:constants

verb:default

verb:encode

verb:max

verb:session

AiProvider

AskError

AskRequest

AskResponse

AskStatus

ImageAttachment

precheck (other)

CliGateDecision

Precheck

How to use it

Module structure

Public surface

Boundary

Where it sits

Dependencies

Feature flags

Targets

Error model

Operational characteristics

Configuration

Related capabilities

Verification

Documentation coverage

Metrics

Generation

`verb:ask`

`verb:busy`

`verb:constants`

`verb:default`

`verb:encode`

`verb:max`

`verb:session`

`AiProvider`

`AskError`

`AskRequest`

`AskResponse`

`AskStatus`

`ImageAttachment`

`precheck (other)`

`CliGateDecision`

`Precheck`