fixed extraction logic

This commit is contained in:
2026-01-18 12:05:36 -05:00
parent 488138a389
commit 1e00372140
8 changed files with 674 additions and 71 deletions

View File

@@ -30,6 +30,14 @@ Forbidden:
* Inferred intent * Inferred intent
* Heuristic task selection * Heuristic task selection
Allowed (explicit exception):
* Autogenerated or inferred selectors, but **only** when they resolve to one of the following concrete patterns:
* `{ kind: "css"; selector: string }`
* `{ kind: "cssAll"; selector: string; index: number }`
* `{ kind: "textScope"; text: string }`
* `{ kind: "anchoredCss"; anchor: { kind: "textScope"; text: string }; selector: string }`
If a change violates this, it is **out of scope**. If a change violates this, it is **out of scope**.
--- ---
@@ -879,4 +887,3 @@ Legacy convenience is not a justification for ambiguity.
> The tool must always make it obvious *what will run*, *with what*, and *why*. > The tool must always make it obvious *what will run*, *with what*, and *why*.
Agents exist to preserve this clarity — not to smooth it away. Agents exist to preserve this clarity — not to smooth it away.

View File

@@ -1,17 +1,4 @@
const DEFAULT_TASKS = [ const DEFAULT_TASKS = [];
{
id: "task-generic-fit",
name: "Generic Fit",
text:
"You should evaluate for my fit to the job. You don't need to suggest interview prep, we'll leave those for later. a bit of tuning to your answers: please keep things more compact, a single section for the evaluation is enough, you don't need to analyze every bullet point in the posting."
},
{
id: "task-ratings-only",
name: "Ratings Only",
text:
"Give ratings out of 10 with headings and do not include any other text.\n\n1. Fit evaluation: my fit to the role.\n2. Company status: how well this company offers career development for me.\n3. Pay: use $25 CAD per hour as a baseline; rate the compensation."
}
];
const DEFAULT_SETTINGS = { const DEFAULT_SETTINGS = {
apiKey: "", apiKey: "",
@@ -25,9 +12,8 @@ const DEFAULT_SETTINGS = {
apiBaseUrl: "https://api.openai.com/v1", apiBaseUrl: "https://api.openai.com/v1",
apiKeyHeader: "Authorization", apiKeyHeader: "Authorization",
apiKeyPrefix: "Bearer ", apiKeyPrefix: "Bearer ",
model: "gpt-4o-mini", model: "gpt-5.2",
systemPrompt: systemPrompt: "",
"You are a precise, honest assistant. Be concise and avoid inventing details, be critical about evaluations. You should put in a small summary of all the sections at the end. You should answer in no longer than 3 sections including the summary. And remember to bold or italicize key points.",
tasks: DEFAULT_TASKS, tasks: DEFAULT_TASKS,
shortcuts: [], shortcuts: [],
theme: "system", theme: "system",

View File

@@ -1,11 +1,12 @@
function findMinimumScope(text) { function findMinimumScope(text) {
if (!text) return null; if (!text) return null;
const normalized = text.trim(); const normalized = normalizeWhitespace(text);
if (!normalized) return null; if (!normalized) return null;
const walker = document.createTreeWalker(document.body, NodeFilter.SHOW_ELEMENT, { const walker = document.createTreeWalker(document.body, NodeFilter.SHOW_ELEMENT, {
acceptNode: (node) => { acceptNode: (node) => {
if (node.innerText.includes(normalized)) { const nodeText = normalizeWhitespace(node.innerText);
if (nodeText.includes(normalized)) {
return NodeFilter.FILTER_ACCEPT; return NodeFilter.FILTER_ACCEPT;
} }
return NodeFilter.FILTER_REJECT; return NodeFilter.FILTER_REJECT;
@@ -22,6 +23,18 @@ function findMinimumScope(text) {
return deepest; return deepest;
} }
function normalizeWhitespace(value) {
return String(value || "")
.replace(/\r?\n/g, " ")
.replace(/\s+/g, " ")
.trim()
.toLowerCase();
}
function isPlainObject(value) {
return Boolean(value && typeof value === "object" && !Array.isArray(value));
}
function escapeSelector(value) { function escapeSelector(value) {
if (window.CSS && typeof CSS.escape === "function") { if (window.CSS && typeof CSS.escape === "function") {
return CSS.escape(value); return CSS.escape(value);
@@ -29,6 +42,259 @@ function escapeSelector(value) {
return String(value).replace(/[^a-zA-Z0-9_-]/g, "\\$&"); return String(value).replace(/[^a-zA-Z0-9_-]/g, "\\$&");
} }
function buildClassSelector(className) {
const parts = String(className || "")
.trim()
.split(/\s+/)
.filter(Boolean);
if (!parts.length) return "";
return parts.map((name) => `.${escapeSelector(name)}`).join("");
}
function inferCssAllTarget(node) {
if (!node || node.nodeType !== 1) return null;
const classList = node.classList ? Array.from(node.classList) : [];
let best = null;
for (const className of classList) {
if (!className) continue;
const matches = Array.from(document.getElementsByClassName(className));
const index = matches.indexOf(node);
if (index < 0) continue;
if (!best || matches.length < best.matches.length) {
best = { className, index, matches };
}
}
if (best) {
return {
kind: "cssAll",
selector: `.${escapeSelector(best.className)}`,
index: best.index
};
}
const className =
typeof node.className === "string" ? node.className.trim() : "";
if (!className) return null;
const selector = buildClassSelector(className);
if (!selector) return null;
const matches = Array.from(document.getElementsByClassName(className));
const index = matches.indexOf(node);
if (index < 0) return null;
return { kind: "cssAll", selector, index };
}
function inferCssTarget(node) {
if (!node || node.nodeType !== 1) return null;
const selector = buildSelector(node);
if (!selector) return null;
return { kind: "css", selector };
}
function inferAnchoredCssTarget(text) {
const trimmed = String(text || "").trim();
if (!trimmed) return null;
return {
kind: "anchoredCss",
anchor: { kind: "textScope", text: trimmed },
selector: ":scope"
};
}
function inferScopeTargets(text, node) {
const candidates = [];
const cssAll = inferCssAllTarget(node);
if (cssAll) candidates.push(cssAll);
const css = inferCssTarget(node);
if (css) candidates.push(css);
const anchoredCss = inferAnchoredCssTarget(text);
if (anchoredCss) candidates.push(anchoredCss);
const trimmed = String(text || "").trim();
if (trimmed) {
candidates.push({ kind: "textScope", text: trimmed });
}
return candidates;
}
function selectInferredTarget(text, node) {
const candidates = inferScopeTargets(text, node);
for (const candidate of candidates) {
const resolved = resolveExtractionTarget(candidate);
if (!resolved.error && resolved.node === node) {
return candidate;
}
}
return null;
}
function findBestScopeCandidate(text) {
const normalized = String(text || "").trim();
if (!normalized) return null;
const walker = document.createTreeWalker(document.body, NodeFilter.SHOW_ELEMENT, {
acceptNode: (node) => {
if (node.innerText.includes(normalized)) {
return NodeFilter.FILTER_ACCEPT;
}
return NodeFilter.FILTER_REJECT;
}
});
let best = null;
let node = walker.nextNode();
while (node) {
if (node !== document.body) {
const cssAll = inferCssAllTarget(node);
if (cssAll) {
const resolved = resolveExtractionTarget(cssAll);
if (!resolved.error && resolved.node === node) {
const matchCount = document.querySelectorAll(cssAll.selector).length;
if (!best || matchCount < best.matchCount) {
best = { node, target: cssAll, matchCount };
}
}
}
}
node = walker.nextNode();
}
return best;
}
function parseLegacySelectorString(value) {
const trimmed = String(value || "").trim();
if (!trimmed) {
return { error: "Missing extraction target." };
}
const classMatch = trimmed.match(
/^(?:document\.)?getElementsByClassName\(\s*(['"])(.+?)\1\s*\)\s*\[\s*(\d+)\s*\]\s*(?:\.innerText\s*)?;?$/i
);
if (classMatch) {
const selector = buildClassSelector(classMatch[2]);
if (!selector) {
return { error: "Missing extraction target." };
}
const index = Number.parseInt(classMatch[3], 10);
if (!Number.isInteger(index) || index < 0) {
return { error: "Invalid index." };
}
return {
target: { kind: "cssAll", selector, index }
};
}
if (trimmed.includes("getElementsByClassName")) {
return { error: "Unsupported extraction target." };
}
return null;
}
function normalizeExtractionTarget(input) {
if (!input) {
return { error: "Missing extraction target." };
}
if (typeof input === "string") {
const parsed = parseLegacySelectorString(input);
if (parsed) {
if (parsed.error) return { error: parsed.error };
return { target: parsed.target };
}
const selector = input.trim();
if (!selector) {
return { error: "Missing extraction target." };
}
return { target: { kind: "css", selector } };
}
if (!isPlainObject(input) || typeof input.kind !== "string") {
return { error: "Missing extraction target." };
}
return { target: input };
}
function resolveExtractionTarget(target) {
if (!target || typeof target !== "object") {
return { error: "Missing extraction target." };
}
if (target.kind === "xpath") {
return { error: "XPath not supported." };
}
if (target.kind === "textScope") {
if (typeof target.text !== "string" || !target.text.trim()) {
return { error: "Missing extraction target." };
}
const node = findMinimumScope(target.text);
if (!node) {
return { error: "Scope not found." };
}
return { node };
}
if (target.kind === "anchoredCss") {
const anchor = target.anchor;
if (
!anchor ||
anchor.kind !== "textScope" ||
typeof anchor.text !== "string" ||
!anchor.text.trim()
) {
return { error: "Missing extraction target." };
}
const anchorNode = findMinimumScope(anchor.text);
if (!anchorNode) {
return { error: "Anchor scope not found." };
}
const selector = target.selector || "";
if (!selector.trim()) {
return { error: "Missing extraction target." };
}
let node = null;
try {
node = anchorNode.querySelector(selector);
} catch {
return { error: "Invalid selector." };
}
if (!node) {
return { error: "Selector matched no elements." };
}
return { node };
}
if (target.kind === "css" || target.kind === "cssAll") {
const selector = target.selector || "";
if (!selector) {
return { error: "Missing extraction target." };
}
if (target.kind === "css") {
let node = null;
try {
node = document.querySelector(selector);
} catch {
return { error: "Invalid selector." };
}
if (!node) {
return { error: "Selector matched no elements." };
}
return { node };
}
const index = target.index;
if (!Number.isInteger(index) || index < 0) {
return { error: "Invalid index." };
}
let nodes = [];
try {
nodes = Array.from(document.querySelectorAll(selector));
} catch {
return { error: "Invalid selector." };
}
if (!nodes.length) {
return { error: "Selector matched no elements." };
}
if (index >= nodes.length) {
return { error: "Index out of bounds." };
}
return { node: nodes[index] };
}
return { error: "Unsupported extraction target." };
}
function buildSelector(node) { function buildSelector(node) {
if (!node || node.nodeType !== 1) return "body"; if (!node || node.nodeType !== 1) return "body";
if (node === document.body) return "body"; if (node === document.body) return "body";
@@ -308,41 +574,59 @@ observer.observe(document.documentElement, { childList: true, subtree: true });
chrome.runtime.onMessage.addListener((message, _sender, sendResponse) => { chrome.runtime.onMessage.addListener((message, _sender, sendResponse) => {
if (!message || typeof message !== "object") return; if (!message || typeof message !== "object") return;
if (message.type === "FIND_SCOPE") { if (message.type === "FIND_SCOPE") {
const node = findMinimumScope(message.text || ""); const rawText = message.text || "";
if (!node) { const baseTarget = { kind: "textScope", text: rawText };
sendResponse({ ok: false, error: "Scope not found." }); const resolved = resolveExtractionTarget(baseTarget);
if (resolved.error) {
sendResponse({ ok: false, error: resolved.error });
return; return;
} }
let effectiveNode = resolved.node;
let responseTarget = selectInferredTarget(rawText, resolved.node) || baseTarget;
if (resolved.node === document.body) {
const scoped = findBestScopeCandidate(rawText);
if (scoped) {
effectiveNode = scoped.node;
responseTarget = scoped.target;
} else if (
responseTarget.kind === "css" &&
responseTarget.selector === "body"
) {
responseTarget = baseTarget;
}
}
sendResponse({ sendResponse({
ok: true, ok: true,
extracted: node.innerText || "", extracted: effectiveNode.innerText || "",
selector: buildSelector(node) target: responseTarget
}); });
return; return;
} }
if (message.type === "EXTRACT_BY_SELECTOR") { if (message.type === "EXTRACT_BY_SELECTOR") {
const selector = message.selector || ""; const { target, error } = normalizeExtractionTarget(
if (!selector) { message.target ?? message.selector
sendResponse({ ok: false, error: "Missing selector." }); );
if (error) {
sendResponse({ ok: false, error });
return; return;
} }
let node = null; const resolved = resolveExtractionTarget(target);
try { if (resolved.error) {
node = document.querySelector(selector); sendResponse({ ok: false, error: resolved.error });
} catch {
sendResponse({ ok: false, error: "Invalid selector." });
return; return;
} }
if (!node) { sendResponse({ ok: true, extracted: resolved.node.innerText || "", target });
sendResponse({ ok: false, error: "Selector not found." });
return;
}
sendResponse({ ok: true, extracted: node.innerText || "", selector });
return; return;
} }
if (message.type === "EXTRACT_FULL") { if (message.type === "EXTRACT_FULL") {
const extracted = document.body?.innerText || ""; const target = { kind: "css", selector: "body" };
sendResponse({ ok: true, extracted, selector: "body" }); const resolved = resolveExtractionTarget(target);
if (resolved.error) {
const extracted = document.body?.innerText || "";
sendResponse({ ok: true, extracted, target });
return;
}
sendResponse({ ok: true, extracted: resolved.node.innerText || "", target });
} }
}); });

View File

@@ -1,7 +1,7 @@
{ {
"manifest_version": 3, "manifest_version": 3,
"name": "SiteCompanion", "name": "SiteCompanion",
"version": "0.4.1", "version": "0.4.2",
"description": "AI companion for site-bound text extraction and tasks.", "description": "AI companion for site-bound text extraction and tasks.",
"permissions": ["storage", "activeTab"], "permissions": ["storage", "activeTab"],
"host_permissions": ["<all_urls>"], "host_permissions": ["<all_urls>"],

View File

@@ -275,6 +275,11 @@ button:active {
color: var(--accent-deep); color: var(--accent-deep);
} }
.helper-text {
font-size: 11px;
color: var(--accent-deep);
}
.output { .output {
margin-top: 8px; margin-top: 8px;
border: 1px dashed var(--border); border: 1px dashed var(--border);

View File

@@ -18,8 +18,10 @@
<div class="state-body"> <div class="state-body">
<p>This site is not recognized. Paste partial text from the page you want to extract:</p> <p>This site is not recognized. Paste partial text from the page you want to extract:</p>
<textarea id="partialTextPaste" rows="4" placeholder="Paste some text here..."></textarea> <textarea id="partialTextPaste" rows="4" placeholder="Paste some text here..."></textarea>
<div id="minimalExtractStatus" class="helper-text hidden"></div>
<div class="row"> <div class="row">
<button id="extractFullBtn" class="accent">Try Extracting Full Text</button> <button id="extractMinimalBtn" class="accent">Try Extracting Minimal</button>
<button id="extractFullBtn" class="ghost">Extract Full Text</button>
</div> </div>
</div> </div>
</section> </section>

View File

@@ -24,6 +24,8 @@ const unknownSiteState = document.getElementById("unknownSiteState");
const extractionReviewState = document.getElementById("extractionReviewState"); const extractionReviewState = document.getElementById("extractionReviewState");
const normalExecutionState = document.getElementById("normalExecutionState"); const normalExecutionState = document.getElementById("normalExecutionState");
const partialTextPaste = document.getElementById("partialTextPaste"); const partialTextPaste = document.getElementById("partialTextPaste");
const minimalExtractStatus = document.getElementById("minimalExtractStatus");
const extractMinimalBtn = document.getElementById("extractMinimalBtn");
const extractFullBtn = document.getElementById("extractFullBtn"); const extractFullBtn = document.getElementById("extractFullBtn");
const extractedPreview = document.getElementById("extractedPreview"); const extractedPreview = document.getElementById("extractedPreview");
const siteNameInput = document.getElementById("siteNameInput"); const siteNameInput = document.getElementById("siteNameInput");
@@ -49,7 +51,7 @@ const state = {
currentPopupState: "unknown", currentPopupState: "unknown",
globalTheme: "system", globalTheme: "system",
forcedTask: null, forcedTask: null,
siteTextSelector: "", siteTextTarget: null,
selectedTaskId: "", selectedTaskId: "",
selectedEnvId: "", selectedEnvId: "",
selectedProfileId: "" selectedProfileId: ""
@@ -68,6 +70,7 @@ async function switchState(stateName) {
} else if (stateName === "normal") { } else if (stateName === "normal") {
normalExecutionState.classList.remove("hidden"); normalExecutionState.classList.remove("hidden");
} }
setMinimalStatus("");
await chrome.storage.local.set({ lastPopupState: stateName }); await chrome.storage.local.set({ lastPopupState: stateName });
} }
@@ -77,7 +80,7 @@ function buildPopupDraft() {
siteText: state.siteText || "", siteText: state.siteText || "",
urlPattern: urlPatternInput?.value?.trim() || "", urlPattern: urlPatternInput?.value?.trim() || "",
siteName: siteNameInput?.value?.trim() || "", siteName: siteNameInput?.value?.trim() || "",
siteTextSelector: state.siteTextSelector || "" siteTextTarget: state.siteTextTarget
}; };
} }
@@ -101,8 +104,10 @@ function applyPopupDraft(draft) {
if (typeof draft.siteName === "string") { if (typeof draft.siteName === "string") {
siteNameInput.value = draft.siteName; siteNameInput.value = draft.siteName;
} }
if (typeof draft.siteTextSelector === "string") { if (draft.siteTextTarget) {
state.siteTextSelector = draft.siteTextSelector; state.siteTextTarget = draft.siteTextTarget;
} else if (typeof draft.siteTextSelector === "string") {
state.siteTextTarget = { kind: "css", selector: draft.siteTextSelector };
} }
} }
@@ -127,6 +132,45 @@ function normalizeName(value) {
return (value || "").trim().toLowerCase(); return (value || "").trim().toLowerCase();
} }
function escapeSelector(value) {
if (window.CSS && typeof CSS.escape === "function") {
return CSS.escape(value);
}
return String(value).replace(/[^a-zA-Z0-9_-]/g, "\\$&");
}
function buildClassSelector(className) {
const parts = String(className || "")
.trim()
.split(/\s+/)
.filter(Boolean);
if (!parts.length) return "";
return parts.map((name) => `.${escapeSelector(name)}`).join("");
}
function parseLegacyDomSelectorString(rawValue) {
const trimmed = String(rawValue || "").trim();
if (!trimmed) return null;
const classMatch = trimmed.match(
/^(?:document\.)?getElementsByClassName\(\s*(['"])(.+?)\1\s*\)\s*\[\s*(\d+)\s*\]\s*(?:\.innerText\s*)?;?$/i
);
if (classMatch) {
const selector = buildClassSelector(classMatch[2]);
if (!selector) {
return { target: null, error: "Missing extraction target." };
}
const index = Number.parseInt(classMatch[3], 10);
if (!Number.isInteger(index) || index < 0) {
return { target: null, error: "Invalid index." };
}
return { target: { kind: "cssAll", selector, index }, error: null };
}
if (trimmed.includes("getElementsByClassName")) {
return { target: null, error: "Unsupported extraction target." };
}
return null;
}
function normalizeConfigList(list) { function normalizeConfigList(list) {
return Array.isArray(list) return Array.isArray(list)
? list.map((item) => ({ ...item, enabled: item.enabled !== false })) ? list.map((item) => ({ ...item, enabled: item.enabled !== false }))
@@ -181,6 +225,26 @@ function resolveEffectiveList(globalItems, workspace, site, listKey, disabledKey
return resolveScopedItems(workspaceEffective, siteItems, siteDisabled); return resolveScopedItems(workspaceEffective, siteItems, siteDisabled);
} }
function normalizeStoredExtractTarget(site) {
if (!site || typeof site !== "object") return null;
const direct = site.extractTarget;
if (direct && typeof direct === "object" && typeof direct.kind === "string") {
return direct;
}
if (typeof direct === "string" && direct.trim()) {
const legacy = parseLegacyDomSelectorString(direct);
if (legacy?.target) return legacy.target;
return { kind: "css", selector: direct.trim() };
}
const legacy = site.extractSelector;
if (typeof legacy === "string" && legacy.trim()) {
const parsedLegacy = parseLegacyDomSelectorString(legacy);
if (parsedLegacy?.target) return parsedLegacy.target;
return { kind: "css", selector: legacy.trim() };
}
return null;
}
function filterApiConfigsForScope(apiConfigs, workspace, site) { function filterApiConfigsForScope(apiConfigs, workspace, site) {
const workspaceDisabled = workspace?.disabledInherited?.apiConfigs || []; const workspaceDisabled = workspace?.disabledInherited?.apiConfigs || [];
const siteDisabled = site?.disabledInherited?.apiConfigs || []; const siteDisabled = site?.disabledInherited?.apiConfigs || [];
@@ -197,7 +261,7 @@ async function detectSite(url) {
const { sites = [], workspaces = [] } = await getStorage(["sites", "workspaces"]); const { sites = [], workspaces = [] } = await getStorage(["sites", "workspaces"]);
const normalizedSites = (Array.isArray(sites) ? sites : []).map((site) => ({ const normalizedSites = (Array.isArray(sites) ? sites : []).map((site) => ({
...site, ...site,
extractSelector: site?.extractSelector || "body" extractTarget: normalizeStoredExtractTarget(site)
})); }));
state.sites = normalizedSites; state.sites = normalizedSites;
state.workspaces = workspaces; state.workspaces = workspaces;
@@ -433,6 +497,12 @@ function setStatus(message) {
statusEl.textContent = message; statusEl.textContent = message;
} }
function setMinimalStatus(message) {
if (!minimalExtractStatus) return;
minimalExtractStatus.textContent = message || "";
minimalExtractStatus.classList.toggle("hidden", !message);
}
function applyTheme(theme) { function applyTheme(theme) {
const value = theme || "system"; const value = theme || "system";
document.documentElement.dataset.theme = value; document.documentElement.dataset.theme = value;
@@ -704,17 +774,24 @@ async function loadConfig() {
const envs = normalizeConfigList(stored.envConfigs); const envs = normalizeConfigList(stored.envConfigs);
const profiles = normalizeConfigList(stored.profiles); const profiles = normalizeConfigList(stored.profiles);
const shortcuts = normalizeConfigList(stored.shortcuts); const shortcuts = normalizeConfigList(stored.shortcuts);
let needsSiteUpdate = false;
const sites = Array.isArray(stored.sites) const sites = Array.isArray(stored.sites)
? stored.sites.map((site) => ({ ? stored.sites.map((site) => {
...site, const target = normalizeStoredExtractTarget(site);
extractSelector: site?.extractSelector || "body" if (site?.extractSelector || typeof site?.extractTarget === "string") {
})) needsSiteUpdate = true;
}
return { ...site, extractTarget: target };
})
: state.sites; : state.sites;
const workspaces = Array.isArray(stored.workspaces) const workspaces = Array.isArray(stored.workspaces)
? stored.workspaces ? stored.workspaces
: state.workspaces; : state.workspaces;
state.sites = sites; state.sites = sites;
state.workspaces = workspaces; state.workspaces = workspaces;
if (needsSiteUpdate) {
await chrome.storage.local.set({ sites });
}
const activeSite = state.currentSite const activeSite = state.currentSite
? sites.find((entry) => entry.id === state.currentSite.id) ? sites.find((entry) => entry.id === state.currentSite.id)
@@ -812,10 +889,14 @@ async function loadTheme() {
async function handleExtract() { async function handleExtract() {
setStatus("Extracting..."); setStatus("Extracting...");
try { try {
const selector = state.currentSite?.extractSelector || "body"; const target = normalizeStoredExtractTarget(state.currentSite);
if (!target) {
setStatus("Missing extraction target.");
return false;
}
const response = await sendToActiveTab({ const response = await sendToActiveTab({
type: "EXTRACT_BY_SELECTOR", type: "EXTRACT_BY_SELECTOR",
selector target
}); });
if (!response?.ok) { if (!response?.ok) {
setStatus(response?.error || "No text detected."); setStatus(response?.error || "No text detected.");
@@ -823,7 +904,7 @@ async function handleExtract() {
} }
state.siteText = response.extracted || ""; state.siteText = response.extracted || "";
state.siteTextSelector = response.selector || selector; state.siteTextTarget = response.target || target;
updateSiteTextCount(); updateSiteTextCount();
updatePromptCount(0); updatePromptCount(0);
setStatus("Text extracted."); setStatus("Text extracted.");
@@ -1045,45 +1126,70 @@ async function fillSiteDefaultsFromTab() {
const tabs = await chrome.tabs.query({ active: true, currentWindow: true }); const tabs = await chrome.tabs.query({ active: true, currentWindow: true });
if (!tabs[0]?.url) return; if (!tabs[0]?.url) return;
const url = new URL(tabs[0].url); const url = new URL(tabs[0].url);
urlPatternInput.value = url.hostname + url.pathname + "*"; urlPatternInput.value = `${url.hostname}/*`;
if (!siteNameInput.value.trim()) { if (!siteNameInput.value.trim()) {
siteNameInput.value = url.hostname; siteNameInput.value = url.hostname;
} }
} }
partialTextPaste.addEventListener("input", async () => { async function runMinimalExtraction(text, minLength = 5) {
const text = partialTextPaste.value.trim(); const trimmed = (text || "").trim();
if (text.length < 5) return; if (trimmed.length < minLength) {
setMinimalStatus("Paste more text to extract.");
return false;
}
setStatus("Finding scope..."); setStatus("Finding scope...");
try { try {
const response = await sendToActiveTab({ type: "FIND_SCOPE", text }); const response = await sendToActiveTab({ type: "FIND_SCOPE", text: trimmed });
if (response?.ok) { if (response?.ok) {
state.siteText = response.extracted; state.siteText = response.extracted;
state.siteTextSelector = response.selector || ""; state.siteTextTarget = response.target || { kind: "textScope", text: trimmed };
extractedPreview.textContent = state.siteText; extractedPreview.textContent = state.siteText;
await fillSiteDefaultsFromTab(); await fillSiteDefaultsFromTab();
switchState("review"); switchState("review");
await persistPopupDraft(); await persistPopupDraft();
setMinimalStatus("");
setStatus("Review extraction."); setStatus("Review extraction.");
return true;
} }
setMinimalStatus(response?.error || "Text could not be matched.");
return false;
} catch (error) { } catch (error) {
setStatus("Error finding scope."); setMinimalStatus(error?.message || "Error finding scope.");
return false;
}
}
partialTextPaste.addEventListener("input", () => {
if (state.currentPopupState === "unknown") {
void persistPopupDraft();
setMinimalStatus("");
} }
}); });
extractMinimalBtn?.addEventListener("click", async () => {
await runMinimalExtraction(partialTextPaste.value, 1);
});
extractFullBtn.addEventListener("click", async () => { extractFullBtn.addEventListener("click", async () => {
setMinimalStatus("");
setStatus("Extracting full text..."); setStatus("Extracting full text...");
try { try {
const response = await sendToActiveTab({ type: "EXTRACT_FULL" }); const response = await sendToActiveTab({
type: "EXTRACT_FULL"
});
if (response?.ok) { if (response?.ok) {
const target = response.target || { kind: "css", selector: "body" };
state.siteText = response.extracted; state.siteText = response.extracted;
state.siteTextSelector = response.selector || "body"; state.siteTextTarget = target;
extractedPreview.textContent = state.siteText; extractedPreview.textContent = state.siteText;
await fillSiteDefaultsFromTab(); await fillSiteDefaultsFromTab();
switchState("review"); switchState("review");
await persistPopupDraft(); await persistPopupDraft();
setStatus("Review extraction."); setStatus("Review extraction.");
} else {
setStatus(response?.error || "Error extracting text.");
} }
} catch (error) { } catch (error) {
setStatus("Error extracting text."); setStatus("Error extracting text.");
@@ -1107,7 +1213,8 @@ retryExtractBtn.addEventListener("click", () => {
urlPatternInput.value = ""; urlPatternInput.value = "";
siteNameInput.value = ""; siteNameInput.value = "";
state.siteText = ""; state.siteText = "";
state.siteTextSelector = ""; state.siteTextTarget = null;
setMinimalStatus("");
void clearPopupDraft(); void clearPopupDraft();
setStatus("Ready."); setStatus("Ready.");
}); });
@@ -1123,6 +1230,10 @@ confirmSiteBtn.addEventListener("click", async () => {
setStatus("Enter a URL pattern."); setStatus("Enter a URL pattern.");
return; return;
} }
if (!state.siteTextTarget) {
setStatus("Missing extraction target.");
return;
}
// AGENTS.md: No URL pattern may be a substring of another. // AGENTS.md: No URL pattern may be a substring of another.
const conflict = state.sites.find(s => s.urlPattern.includes(pattern) || pattern.includes(s.urlPattern)); const conflict = state.sites.find(s => s.urlPattern.includes(pattern) || pattern.includes(s.urlPattern));
@@ -1136,7 +1247,7 @@ confirmSiteBtn.addEventListener("click", async () => {
name, name,
urlPattern: pattern, urlPattern: pattern,
workspaceId: "global", // Default to global for now workspaceId: "global", // Default to global for now
extractSelector: state.siteTextSelector || "body" extractTarget: state.siteTextTarget
}; };
state.sites.push(newSite); state.sites.push(newSite);

View File

@@ -29,11 +29,205 @@ const OPENAI_DEFAULTS = {
apiKeyHeader: "Authorization", apiKeyHeader: "Authorization",
apiKeyPrefix: "Bearer " apiKeyPrefix: "Bearer "
}; };
const DEFAULT_MODEL = "gpt-4o-mini"; const DEFAULT_MODEL = "gpt-5.2";
const DEFAULT_SYSTEM_PROMPT = const DEFAULT_SYSTEM_PROMPT = "";
"You are a precise, honest assistant. Be concise and avoid inventing details, be critical about evaluations. You should put in a small summary of all the sections at the end. You should answer in no longer than 3 sections including the summary. And remember to bold or italicize key points.";
const SIDEBAR_WIDTH_KEY = "sidebarWidth"; const SIDEBAR_WIDTH_KEY = "sidebarWidth";
function isPlainObject(value) {
return Boolean(value && typeof value === "object" && !Array.isArray(value));
}
function escapeSelector(value) {
if (window.CSS && typeof CSS.escape === "function") {
return CSS.escape(value);
}
return String(value).replace(/[^a-zA-Z0-9_-]/g, "\\$&");
}
function buildClassSelector(className) {
const parts = String(className || "")
.trim()
.split(/\s+/)
.filter(Boolean);
if (!parts.length) return "";
return parts.map((name) => `.${escapeSelector(name)}`).join("");
}
function parseLegacyDomSelectorString(rawValue) {
const trimmed = String(rawValue || "").trim();
if (!trimmed) return null;
const classMatch = trimmed.match(
/^(?:document\.)?getElementsByClassName\(\s*(['"])(.+?)\1\s*\)\s*\[\s*(\d+)\s*\]\s*(?:\.innerText\s*)?;?$/i
);
if (classMatch) {
const selector = buildClassSelector(classMatch[2]);
if (!selector) {
return { target: null, error: "Missing extraction target." };
}
const index = Number.parseInt(classMatch[3], 10);
if (!Number.isInteger(index) || index < 0) {
return { target: null, error: "Invalid index." };
}
return { target: { kind: "cssAll", selector, index }, error: null };
}
if (trimmed.includes("getElementsByClassName")) {
return { target: null, error: "Unsupported extraction target." };
}
return null;
}
function parseLooseJsonInput(rawValue) {
const trimmed = String(rawValue || "").trim();
if (!trimmed.startsWith("{")) return null;
let normalized = trimmed;
normalized = normalized.replace(
/([{,]\s*)([A-Za-z_][A-Za-z0-9_]*)(\s*:)/g,
'$1"$2"$3'
);
normalized = normalized.replace(
/'([^'\\]*(?:\\.[^'\\]*)*)'/g,
(_match, value) => `"${value.replace(/"/g, '\\"')}"`
);
return normalized;
}
function normalizeExtractionTargetValue(value) {
if (typeof value === "string") {
const legacy = parseLegacyDomSelectorString(value);
if (legacy) {
return legacy.target;
}
const trimmed = value.trim();
return trimmed ? { kind: "css", selector: trimmed } : null;
}
if (isPlainObject(value) && typeof value.kind === "string") {
return value;
}
return null;
}
function serializeExtractionTarget(target) {
if (!target) return "";
if (typeof target === "string") {
const legacy = parseLegacyDomSelectorString(target);
if (legacy?.target) return JSON.stringify(legacy.target);
const trimmed = target.trim();
if (!trimmed) return "";
return JSON.stringify({ kind: "css", selector: trimmed });
}
if (isPlainObject(target) && typeof target.kind === "string") {
return JSON.stringify(target);
}
return "";
}
function validateExtractionTarget(target) {
if (!target || typeof target !== "object") {
return "Missing extraction target.";
}
if (target.kind === "xpath") {
return "XPath not supported.";
}
if (target.kind === "css") {
return typeof target.selector === "string" && target.selector.trim()
? null
: "Missing extraction target.";
}
if (target.kind === "cssAll") {
if (typeof target.selector !== "string" || !target.selector.trim()) {
return "Missing extraction target.";
}
if (!Number.isInteger(target.index) || target.index < 0) {
return "Invalid index.";
}
return null;
}
if (target.kind === "textScope") {
return typeof target.text === "string" && target.text.trim()
? null
: "Missing extraction target.";
}
if (target.kind === "anchoredCss") {
const anchor = target.anchor;
if (!anchor || anchor.kind !== "textScope") {
return "Invalid anchor target.";
}
if (typeof anchor.text !== "string" || !anchor.text.trim()) {
return "Missing extraction target.";
}
if (typeof target.selector !== "string" || !target.selector.trim()) {
return "Missing extraction target.";
}
return null;
}
return "Unsupported extraction target.";
}
function parseExtractionTargetInput(rawValue) {
const trimmed = (rawValue || "").trim();
if (!trimmed) {
return { target: null, error: "Missing extraction target." };
}
const legacy = parseLegacyDomSelectorString(trimmed);
if (legacy) {
if (legacy.error) {
return { target: null, error: legacy.error };
}
const error = validateExtractionTarget(legacy.target);
return { target: legacy.target, error };
}
if (trimmed.startsWith("textScope:")) {
const text = trimmed.slice("textScope:".length).trim();
const target = { kind: "textScope", text };
const error = validateExtractionTarget(target);
return { target, error };
}
let target = null;
if (trimmed.startsWith("{")) {
try {
const parsed = JSON.parse(trimmed);
target = normalizeExtractionTargetValue(parsed);
} catch {
const normalized = parseLooseJsonInput(trimmed);
if (!normalized) {
return { target: null, error: "Invalid extraction target JSON." };
}
try {
const parsed = JSON.parse(normalized);
target = normalizeExtractionTargetValue(parsed);
} catch {
return { target: null, error: "Invalid extraction target JSON." };
}
}
} else {
target = { kind: "css", selector: trimmed };
}
if (!target) {
return { target: null, error: "Invalid extraction target." };
}
const error = validateExtractionTarget(target);
return { target, error };
}
function normalizeStoredExtractionTarget(site) {
const normalized = normalizeExtractionTargetValue(site?.extractTarget);
if (normalized) {
const changed = typeof site?.extractTarget === "string";
return { target: normalized, changed };
}
if (typeof site?.extractSelector === "string" && site.extractSelector.trim()) {
const legacy = parseLegacyDomSelectorString(site.extractSelector);
if (legacy?.target) {
return { target: legacy.target, changed: true };
}
return {
target: { kind: "css", selector: site.extractSelector.trim() },
changed: true
};
}
return { target: null, changed: false };
}
function getSidebarWidthLimits() { function getSidebarWidthLimits() {
const min = 160; const min = 160;
const max = Math.max(min, Math.min(360, window.innerWidth - 240)); const max = Math.max(min, Math.min(360, window.innerWidth - 240));
@@ -2481,6 +2675,7 @@ function collectSites() {
const patternInput = card.querySelector(".site-pattern"); const patternInput = card.querySelector(".site-pattern");
const workspaceSelect = card.querySelector(".site-workspace"); const workspaceSelect = card.querySelector(".site-workspace");
const extractInput = card.querySelector(".site-extract-selector"); const extractInput = card.querySelector(".site-extract-selector");
const parsedTarget = parseExtractionTargetInput(extractInput?.value || "");
const themeSelect = card.querySelector(".appearance-theme"); const themeSelect = card.querySelector(".appearance-theme");
const toolbarSelect = card.querySelector(".appearance-toolbar-position"); const toolbarSelect = card.querySelector(".appearance-toolbar-position");
const envsContainer = card.querySelector(".site-envs"); const envsContainer = card.querySelector(".site-envs");
@@ -2497,7 +2692,7 @@ function collectSites() {
name: (nameInput?.value || "").trim(), name: (nameInput?.value || "").trim(),
urlPattern: (patternInput?.value || "").trim(), urlPattern: (patternInput?.value || "").trim(),
workspaceId: workspaceSelect?.value || "global", workspaceId: workspaceSelect?.value || "global",
extractSelector: (extractInput?.value || "").trim(), extractTarget: parsedTarget.target,
theme: themeSelect?.value || "inherit", theme: themeSelect?.value || "inherit",
toolbarPosition: toolbarSelect?.value || "inherit", toolbarPosition: toolbarSelect?.value || "inherit",
envConfigs: envsContainer ? collectEnvConfigs(envsContainer) : [], envConfigs: envsContainer ? collectEnvConfigs(envsContainer) : [],
@@ -2619,7 +2814,7 @@ function buildSiteCard(site, allWorkspaces = []) {
extractLabel.textContent = "Site Text Selector"; extractLabel.textContent = "Site Text Selector";
const extractInput = document.createElement("input"); const extractInput = document.createElement("input");
extractInput.type = "text"; extractInput.type = "text";
extractInput.value = site.extractSelector || ""; extractInput.value = serializeExtractionTarget(site.extractTarget);
extractInput.className = "site-extract-selector"; extractInput.className = "site-extract-selector";
extractInput.placeholder = "body"; extractInput.placeholder = "body";
extractInput.addEventListener("input", () => { extractInput.addEventListener("input", () => {
@@ -3203,6 +3398,11 @@ function updateSidebarErrors() {
".shortcut-name", ".shortcut-name",
`${label} shortcuts` `${label} shortcuts`
); );
const extractInput = card.querySelector(".site-extract-selector");
const { error } = parseExtractionTargetInput(extractInput?.value || "");
if (error) {
errors.push(`${label} site text selector: ${error}`);
}
}); });
checkNameInputs(sitesContainer, ".site-name", "Sites"); checkNameInputs(sitesContainer, ".site-name", "Sites");
@@ -3408,13 +3608,18 @@ async function loadSettings() {
} }
if (Array.isArray(sites)) { if (Array.isArray(sites)) {
let needsSiteUpdate = false;
sites = sites.map((site) => { sites = sites.map((site) => {
if (!site || typeof site !== "object") return site; if (!site || typeof site !== "object") return site;
const normalizedTarget = normalizeStoredExtractionTarget(site);
if (normalizedTarget.changed) {
needsSiteUpdate = true;
}
return { return {
...site, ...site,
name: site.name || site.urlPattern || "", name: site.name || site.urlPattern || "",
workspaceId: site.workspaceId || "global", workspaceId: site.workspaceId || "global",
extractSelector: typeof site.extractSelector === "string" ? site.extractSelector : "", extractTarget: normalizedTarget.target,
theme: site.theme || "inherit", theme: site.theme || "inherit",
toolbarPosition: site.toolbarPosition || "inherit", toolbarPosition: site.toolbarPosition || "inherit",
envConfigs: normalizeConfigList(site.envConfigs), envConfigs: normalizeConfigList(site.envConfigs),
@@ -3424,6 +3629,9 @@ async function loadSettings() {
disabledInherited: normalizeDisabledInherited(site.disabledInherited) disabledInherited: normalizeDisabledInherited(site.disabledInherited)
}; };
}); });
if (needsSiteUpdate) {
await chrome.storage.local.set({ sites });
}
} }
// Load basic resources first so they are available for shortcuts/workspaces // Load basic resources first so they are available for shortcuts/workspaces