chore(deps): make tokio optional

.github/workflows/ci.yml

@@ -89,6 +89,14 @@ jobs:
             echo "HTTP client leaked into the docx,spreadsheet,markdown,email feature set"
             exit 1
           fi
+          if cargo tree -p markkit --no-default-features --features docx,spreadsheet -e normal | grep -E 'tokio'; then
+            echo "tokio leaked into the docx,spreadsheet feature set"
+            exit 1
+          fi
+          if cargo tree -p markkit -e normal | grep -E 'tokio'; then
+            echo "tokio leaked into the default feature set"
+            exit 1
+          fi
       - name: Run examples
         run: |
           for ex in basic custom_selectors configuration ir_access transforms; do

AGENTS.md

@@ -98,7 +98,7 @@ just pre-pr        # Pre-PR checks
 - Stable Rust, pinned in `rust-toolchain.toml`
 - `cargo fmt`, `cargo clippy --workspace --all-targets -- -D warnings`
 - Licenses/advisories: `cargo deny check` (`deny.toml`), `cargo audit`
-- Format deps stay feature-gated; `http_client` opt-in (never default); default and `docx,spreadsheet` builds must not link reqwest/hyper (CI checks)
+- Format deps stay feature-gated; `http_client` opt-in (never default); default and `docx,spreadsheet` builds must not link reqwest/hyper or tokio (CI checks)
 
 ### Pre-PR Checklist
 

CHANGELOG.md

@@ -157,6 +157,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
   collapsing).
 - Spreadsheets make a third fewer allocations per cell (XLSX values are formatted without an
   intermediate copy and trimmed in place).
+- Dependencies: `tokio` is optional, pulled in only by `http_client` and `tool`, so default and
+  `docx,spreadsheet` builds no longer link it. CI checks both trees. `serde_json` stays required
+  (always-on `.ipynb` input).
 - **Breaking**: limit fields moved from `MarkkitBuilder` to `MarkkitBuilder::limits`
   (`config().max_input_size` → `config().limits.max_input_size`); `max_processing_secs` field is
   now `limits.timeout: Duration`. Builder setters are unchanged, plus new `timeout`,

README.md

@@ -209,7 +209,8 @@ let markkit = Markkit::builder()
 | `tool` | No | `markkit::tool::MarkkitTool` LLM agent tool (Everruns toolkit contract) |
 | `logging` | No | Reserved: pulls in `tracing`, emits no events yet |
 
-`http_client` is opt-in: the default build has no HTTP/TLS stack in the dependency tree.
+`http_client` is opt-in: the default build has no HTTP/TLS stack and no tokio in the dependency
+tree.
 Enable it to get `extract_url` (the CLI and the Python/Node bindings ship with it on):
 
 ```toml

crates/markkit/Cargo.toml

@@ -23,17 +23,17 @@ spreadsheet = ["dep:calamine", "dep:zip", "dep:quick-xml", "dep:cfb"]
 markdown = ["dep:pulldown-cmark"]
 email = ["dep:mail-parser"]
 # Opt-in: the default build has no network stack (no reqwest/hyper/TLS).
-http_client = ["dep:reqwest", "tokio/net"]
+http_client = ["dep:reqwest", "dep:tokio", "tokio/net"]
 logging = ["dep:tracing"]
 # LLM tool contract (`markkit::tool`): Everruns toolkit-library shape.
-tool = ["dep:tower", "dep:futures-core", "dep:base64"]
+tool = ["dep:tower", "dep:futures-core", "dep:base64", "dep:tokio", "tokio/sync"]
 # JSON Schema for the IR (`markkit::ir::json_schema()`)
 json-schema = ["dep:schemars"]
 all_formats = ["html", "pdf", "docx", "pptx", "spreadsheet", "markdown", "email"]
 
 [dependencies]
-# Async
-tokio = { workspace = true, features = ["sync", "io-util"] }
+# Async (optional): only URL fetching and the tool's output channel need tokio.
+tokio = { workspace = true, optional = true }
 
 # Error handling
 thiserror.workspace = true

knowledge/foundations/architecture.md

@@ -21,7 +21,7 @@ It follows a pipeline architecture: **detect → parse → normalize → convert
 2. **Streaming where possible**: Process large documents without loading entirely into memory
 3. **Configurable output**: Users control what gets included/excluded
 4. **Safe by default**: All input is treated as untrusted; no code execution
-5. **Feature-gated formats**: Only compile what you need
+5. **Feature-gated formats**: Only compile what you need. `tokio` is optional (`http_client`, `tool`). `serde_json` stays required: always-on Jupyter notebook input parses JSON
 
 ## Pipeline
 

knowledge/log.md

@@ -10,6 +10,7 @@
 * **Over-escaping fix**: line-start `#`, `+` and ordered markers are escaped only when followed by whitespace or the end of the text run, so `4.6 out of 5` no longer renders `4\.6` (eval gap). `> - =` stay always escaped. See [Threat Model](security/threat-model.md) T1, [Evaluation Harness](operations/eval.md).
 * **MCP server**: `markkit mcp` serves `MarkkitTool` over MCP stdio (initialize, ping, tools/list, tools/call); line size capped from `--max-size`, tool failures as `isError` results. New [MCP Server](integrations/mcp-server.md) concept; MCP messages row in the [Threat Model](security/threat-model.md).
 * **Nesting cost**: HTML parsing was quadratic in nesting depth (html5ever) and exponential in nested tables (descendant row selection duplicated inner rows); text flattened past `max_nesting_depth` leaked `<script>`/`<style>` bodies. Fixed with a pre-scan nesting cap, direct-structure table rows and a non-content-aware flattener; new invariant S21. Found by `markkit-bench`. See [HTML Extraction](formats/html-extraction.md) and [Threat Model](security/threat-model.md) T2c.
+* **Optional tokio (#11)**: `tokio` only via `http_client` / `tool`; CI asserts it is not in the default or `docx,spreadsheet` tree. `serde_json` stays required because notebook input is always on. See [Markkit Architecture](foundations/architecture.md).
 * **Spreadsheet follow-ups (#11)**: single-row sheets render as data under column letters; XLSX/ODS/XLS titles come from document properties; XLSB content type no longer maps to XLSX; XLSX without `_rels/.rels` parses; trailing-newline output documented as intended; release profile switched to `panic = "unwind"` so `catch_unwind` is effective in markkit's own binaries. See [Format Support](formats/format-support.md) and [Threat Model](security/threat-model.md) T6c.
 * **Plain text and Markdown input**: text is always on and goes through the IR; Markdown (`markdown` feature) is re-parsed with pulldown-cmark rather than passed through, so raw HTML never reaches output. Neither is sniffed from bytes. Shares the call's `Limits` budget (timeout, nodes, nesting depth). New threat T7 and invariant S18. See [Format Support](formats/format-support.md) and [Threat Model](security/threat-model.md).
 * **Eval harness on mira**: `markkit-eval` is now a mira study (modeled on bashkit-eval). `markkit_extract` scores markkit's HTML output on ten realistic pages with typed deterministic checks and one scorer per metric; a `mode` axis compares defaults with `main_content` (73.0% vs 86.8% mean, noise 34% vs 82%); dated results committed in `crates/markkit-eval/results/`; gaps recorded (`main_content` drops the forum question, `4.6` over-escaped as `4\.6`, `tfoot`/`caption` dropped, edit links in headings). See [Evaluation Harness](operations/eval.md).