pyinterviewbot-ai 0.1.0__tar.gz

pyinterviewbot_ai-0.1.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

pyinterviewbot_ai-0.1.0/PKG-INFO

@@ -0,0 +1,25 @@
+Metadata-Version: 2.4
+Name: pyinterviewbot-ai
+Version: 0.1.0
+Summary: Local ASR/LLM/TTS + evidence-engine orchestration service for PyInterviewBot
+Author: Georgi Mammen Mullassery
+Author-email: Georgi Mammen Mullassery <mullassery@gmail.com>
+License-Expression: Apache-2.0
+License-File: LICENSE
+Requires-Dist: aiosqlite>=0.22.1
+Requires-Dist: fastapi>=0.141.1
+Requires-Dist: faster-whisper>=1.2.1
+Requires-Dist: ollama>=0.6.2
+Requires-Dist: piper-tts>=1.8.0
+Requires-Dist: pydantic>=2.13.5
+Requires-Dist: python-multipart>=0.0.32
+Requires-Dist: uvicorn[standard]>=0.53.0
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+
+# pyinterviewbot-ai
+
+The ASR/LLM/TTS orchestration + evidence-engine service for
+[PyInterviewBot](../README.md). See the root README for setup/run
+instructions and [`../docs/ARCHITECTURE.md`](../docs/ARCHITECTURE.md) for
+the design.

pyinterviewbot_ai-0.1.0/README.md

@@ -0,0 +1,6 @@
+# pyinterviewbot-ai
+
+The ASR/LLM/TTS orchestration + evidence-engine service for
+[PyInterviewBot](../README.md). See the root README for setup/run
+instructions and [`../docs/ARCHITECTURE.md`](../docs/ARCHITECTURE.md) for
+the design.

pyinterviewbot_ai-0.1.0/pyproject.toml

@@ -0,0 +1,36 @@
+[project]
+name = "pyinterviewbot-ai"
+version = "0.1.0"
+description = "Local ASR/LLM/TTS + evidence-engine orchestration service for PyInterviewBot"
+readme = "README.md"
+license = "Apache-2.0"
+license-files = ["LICENSE"]
+authors = [
+    { name = "Georgi Mammen Mullassery", email = "mullassery@gmail.com" }
+]
+requires-python = ">=3.11"
+dependencies = [
+    "aiosqlite>=0.22.1",
+    "fastapi>=0.141.1",
+    "faster-whisper>=1.2.1",
+    "ollama>=0.6.2",
+    "piper-tts>=1.8.0",
+    "pydantic>=2.13.5",
+    "python-multipart>=0.0.32",
+    "uvicorn[standard]>=0.53.0",
+]
+
+[project.scripts]
+pyinterviewbot-ai = "pyinterviewbot_ai.main:main"
+
+[build-system]
+requires = ["uv_build>=0.11.32,<0.12.0"]
+build-backend = "uv_build"
+
+[dependency-groups]
+dev = [
+    "httpx>=0.28.1",
+    "pytest>=9.1.1",
+    "pytest-asyncio>=1.4.0",
+    "websockets>=17.1",
+]

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/agent.py

@@ -0,0 +1,152 @@
+"""The interview agent: adaptive question generation + evidence extraction
+(spec §13/§14/§30). One LLM call per turn produces both — the gateway only
+ever sees the resulting question text and a completion flag; everything
+else here stays server-side.
+"""
+
+from __future__ import annotations
+
+import os
+from uuid import UUID, uuid4
+
+from . import db, llm
+from .evidence import EvidenceGraphState
+from .models import TurnResponse
+from .resume import seed_claim
+
+MAX_TURNS = int(os.environ.get("PYINTERVIEWBOT_MAX_TURNS", "6"))
+
+FALLBACK_CLOSING_REMARK = (
+    "That's a great place to stop — thank you for walking me through all of that in such detail."
+)
+
+SYSTEM_PROMPT = """You are a senior technical interviewer conducting a live spoken interview.
+Your tone is professional, calm, intelligent, neutral, and conversational — never a checklist,
+never a quiz show. The candidate cannot see anything you're thinking; they only hear what you say.
+
+You are assessing exactly one competency right now: RAG (Retrieval-Augmented Generation)
+Architecture, grounded in a specific claim from the candidate's resume. Your job is to listen,
+understand, and adaptively probe until you have real evidence — not to work through a fixed
+script.
+
+Evidence nodes for this competency (do not mention these names or ids to the candidate):
+- retrieval_design: how retrieval is designed — what's fetched vs. handled directly by the model
+- relevance_handling: what happens when retrieval returns irrelevant or low-quality documents
+- production_scaling: how the system behaves and is investigated under a large traffic/latency increase
+
+Rules:
+1. Ask exactly ONE natural question or statement per turn.
+2. Never reveal scores, evidence status, confidence, or your internal reasoning to the candidate.
+3. Ground every question in what the candidate actually just said — reference specifics.
+4. Choose a follow-up strategy that fits what's missing: clarification ("when you say X, roughly
+   what scale?"), ownership ("what part did you personally build?"), technical depth ("how did you
+   handle Y?"), trade-off ("why that approach over the alternative?"), production ("what happened
+   once this was live?"), failure ("what was the most serious failure?"), evidence ("walk me
+   through a specific example"), or challenge ("what if load increased tenfold?").
+5. Do not penalize the candidate for topics outside this competency. Only probe RAG architecture.
+6. Once retrieval_design and relevance_handling both have real evidence, transition to
+   production_scaling by posing it as a hypothetical scenario (e.g. "Let's shift into a scenario:
+   traffic just increased twenty-fold and latency has spiked. Walk me through what you'd
+   investigate first."), not a plain question.
+7. Once all three nodes have real evidence (not "not_addressed"), wrap up warmly in one or two
+   sentences, thank them, and set is_complete to true. Do not ask a further question in that turn.
+8. Always fill in evidence_updates for any node the candidate's last answer gave you signal on,
+   using status: "confirmed" (clear, specific, credible), "weak" (vague/generic/hand-wavy),
+   "contradicted" (conflicts with something said earlier), or leave a node out of
+   evidence_updates entirely if this turn didn't touch it.
+
+Respond only with the structured JSON you've been constrained to produce."""
+
+
+def _opening_prompt(claim_text: str) -> str:
+    return (
+        f'The candidate\'s resume includes this claim: "{claim_text}"\n\n'
+        "This is the very first turn. Open the interview by warmly referencing this claim and "
+        "asking them to walk you through the architecture. Leave evidence_updates empty and "
+        "is_complete false."
+    )
+
+
+def _turn_prompt(
+    claim_text: str,
+    graph: EvidenceGraphState,
+    history: list[tuple[str, str]],
+    latest_answer: str,
+    force_conclude: bool,
+) -> str:
+    transcript = "\n".join(f"{'Interviewer' if s == 'ai' else 'Candidate'}: {t}" for s, t in history)
+    closing_instruction = (
+        "\n\nThis is the FINAL turn regardless of evidence coverage — the interview time is up. "
+        "question_text must NOT be a question — it must be 1-2 warm closing sentences thanking the "
+        "candidate and telling them the interview is complete. question_text must not be empty. "
+        "Set is_complete to true."
+        if force_conclude
+        else ""
+    )
+    return (
+        f'Resume claim under discussion: "{claim_text}"\n\n'
+        f"Evidence gathered so far:\n{graph.summary_for_prompt()}\n\n"
+        f"Conversation so far:\n{transcript}\n\n"
+        f"Candidate's latest answer: \"{latest_answer}\"\n\n"
+        "Decide evidence_updates from that answer, then produce your next question per the rules."
+        f"{closing_instruction}"
+    )
+
+
+async def _load_graph(session_id: UUID) -> EvidenceGraphState:
+    graph = EvidenceGraphState()
+    for record in await db.get_evidence(session_id):
+        graph.apply(record.node, record.status, record.note)
+    return graph
+
+
+async def start_session() -> tuple[UUID, str]:
+    session_id = uuid4()
+    claim = seed_claim()
+    await db.create_session(session_id, claim)
+
+    decision = await llm.decide_turn(SYSTEM_PROMPT, _opening_prompt(claim.text))
+
+    await db.append_transcript(session_id, 0, "ai", decision.question_text)
+    await db.set_turn_count(session_id, 0)
+
+    return session_id, decision.question_text
+
+
+async def take_turn(session_id: UUID, transcript: str) -> TurnResponse:
+    row = await db.get_session_row(session_id)
+    if row is None:
+        raise ValueError(f"unknown session {session_id}")
+
+    turn_number = row["turn_count"] + 1
+    claim_text = row["resume_claim"]
+
+    await db.append_transcript(session_id, turn_number, "candidate", transcript)
+
+    graph = await _load_graph(session_id)
+    history = [(s.speaker, s.text) for s in await db.get_transcript(session_id) if s.turn_number < turn_number]
+
+    force_conclude = turn_number >= MAX_TURNS
+    decision = await llm.decide_turn(
+        SYSTEM_PROMPT, _turn_prompt(claim_text, graph, history, transcript, force_conclude)
+    )
+
+    for update in decision.evidence_updates:
+        graph.apply(update.node, update.status, update.note)
+        await db.record_evidence(session_id, update.node, update.status, update.note, turn_number)
+
+    is_complete = decision.is_complete or force_conclude
+    ai_text = decision.question_text.strip()
+    if not ai_text:
+        # Defensive fallback, same spirit as the gateway's ASR-failure
+        # fallback (spec §28) — the primary path is always real LLM output;
+        # this only fires if the model returns nothing for the closing turn.
+        ai_text = FALLBACK_CLOSING_REMARK if is_complete else "Could you tell me a bit more about that?"
+
+    await db.append_transcript(session_id, turn_number, "ai", ai_text)
+    await db.set_turn_count(session_id, turn_number)
+
+    if is_complete:
+        await db.mark_complete(session_id)
+
+    return TurnResponse(ai_text=ai_text, is_complete=is_complete)

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/asr.py

@@ -0,0 +1,47 @@
+"""ASR provider (spec §31): local faster-whisper. Per-turn transcription of
+a complete VAD-delimited utterance — not token-level streaming partials.
+See docs/ARCHITECTURE.md for why that's an intentional MVP simplification
+of the §11/§27 streaming pipeline (faster-whisper has no native incremental
+decode mode; a real streaming-partials setup would periodically re-decode a
+growing buffer, which added complexity this slice skips since nothing
+candidate-facing depends on partial transcripts anyway).
+"""
+
+from __future__ import annotations
+
+import asyncio
+import os
+
+import numpy as np
+from faster_whisper import WhisperModel
+
+MODEL_SIZE = os.environ.get("WHISPER_MODEL", "small.en")
+DEVICE = os.environ.get("WHISPER_DEVICE", "cpu")
+COMPUTE_TYPE = os.environ.get("WHISPER_COMPUTE_TYPE", "int8")
+
+_model: WhisperModel | None = None
+
+
+def _get_model() -> WhisperModel:
+    global _model
+    if _model is None:
+        _model = WhisperModel(MODEL_SIZE, device=DEVICE, compute_type=COMPUTE_TYPE)
+    return _model
+
+
+def _pcm16_bytes_to_float32(pcm16: bytes) -> np.ndarray:
+    samples = np.frombuffer(pcm16, dtype="<i2")
+    return samples.astype(np.float32) / 32768.0
+
+
+def _transcribe_sync(pcm16_16khz: bytes) -> str:
+    audio = _pcm16_bytes_to_float32(pcm16_16khz)
+    if audio.size == 0:
+        return ""
+    segments, _info = _get_model().transcribe(audio, language="en", vad_filter=False)
+    return " ".join(segment.text.strip() for segment in segments).strip()
+
+
+async def transcribe(pcm16_16khz: bytes) -> str:
+    # faster-whisper is synchronous/CPU-bound; keep it off the event loop.
+    return await asyncio.to_thread(_transcribe_sync, pcm16_16khz)

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/db.py

@@ -0,0 +1,173 @@
+"""SQLite persistence — the system of record for interview domain data
+(spec §34: state must not live only inside the LLM context). One file per
+deployment; path configurable via PYINTERVIEWBOT_DB_PATH.
+"""
+
+from __future__ import annotations
+
+import os
+from contextlib import asynccontextmanager
+from datetime import datetime, timezone
+from uuid import UUID
+
+import aiosqlite
+
+from .evidence import COMPETENCY
+from .models import EvidenceRecord, EvidenceStatus, SessionListItem, TranscriptSegment
+from .resume import ResumeClaim
+
+DB_PATH = os.environ.get("PYINTERVIEWBOT_DB_PATH", "data/pyinterviewbot.db")
+
+SCHEMA = """
+CREATE TABLE IF NOT EXISTS sessions (
+    id TEXT PRIMARY KEY,
+    created_at TEXT NOT NULL,
+    resume_claim TEXT NOT NULL,
+    turn_count INTEGER NOT NULL DEFAULT 0,
+    is_complete INTEGER NOT NULL DEFAULT 0
+);
+
+CREATE TABLE IF NOT EXISTS transcript_segments (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    session_id TEXT NOT NULL REFERENCES sessions(id),
+    turn_number INTEGER NOT NULL,
+    speaker TEXT NOT NULL,
+    text TEXT NOT NULL,
+    created_at TEXT NOT NULL
+);
+
+CREATE TABLE IF NOT EXISTS evidence (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    session_id TEXT NOT NULL REFERENCES sessions(id),
+    competency TEXT NOT NULL,
+    node TEXT NOT NULL,
+    status TEXT NOT NULL,
+    note TEXT NOT NULL,
+    turn_number INTEGER NOT NULL,
+    created_at TEXT NOT NULL
+);
+"""
+
+
+async def init_db() -> None:
+    os.makedirs(os.path.dirname(DB_PATH) or ".", exist_ok=True)
+    async with aiosqlite.connect(DB_PATH) as db:
+        await db.executescript(SCHEMA)
+        await db.commit()
+
+
+@asynccontextmanager
+async def connect():
+    db = await aiosqlite.connect(DB_PATH)
+    try:
+        yield db
+    finally:
+        await db.close()
+
+
+def _now() -> str:
+    return datetime.now(timezone.utc).isoformat()
+
+
+async def create_session(session_id: UUID, claim: ResumeClaim) -> None:
+    async with connect() as db:
+        await db.execute(
+            "INSERT INTO sessions (id, created_at, resume_claim) VALUES (?, ?, ?)",
+            (str(session_id), _now(), claim.text),
+        )
+        await db.commit()
+
+
+async def append_transcript(session_id: UUID, turn_number: int, speaker: str, text: str) -> None:
+    async with connect() as db:
+        await db.execute(
+            "INSERT INTO transcript_segments (session_id, turn_number, speaker, text, created_at) "
+            "VALUES (?, ?, ?, ?, ?)",
+            (str(session_id), turn_number, speaker, text, _now()),
+        )
+        await db.commit()
+
+
+async def record_evidence(session_id: UUID, node: str, status: EvidenceStatus, note: str, turn_number: int) -> None:
+    async with connect() as db:
+        await db.execute(
+            "INSERT INTO evidence (session_id, competency, node, status, note, turn_number, created_at) "
+            "VALUES (?, ?, ?, ?, ?, ?, ?)",
+            (str(session_id), COMPETENCY, node, status, note, turn_number, _now()),
+        )
+        await db.commit()
+
+
+async def set_turn_count(session_id: UUID, turn_count: int) -> None:
+    async with connect() as db:
+        await db.execute(
+            "UPDATE sessions SET turn_count = ? WHERE id = ?",
+            (turn_count, str(session_id)),
+        )
+        await db.commit()
+
+
+async def mark_complete(session_id: UUID) -> None:
+    async with connect() as db:
+        await db.execute("UPDATE sessions SET is_complete = 1 WHERE id = ?", (str(session_id),))
+        await db.commit()
+
+
+async def list_sessions() -> list[SessionListItem]:
+    async with connect() as db:
+        db.row_factory = aiosqlite.Row
+        cursor = await db.execute(
+            "SELECT id, created_at, resume_claim, is_complete, turn_count FROM sessions "
+            "ORDER BY created_at DESC LIMIT 50"
+        )
+        rows = await cursor.fetchall()
+        return [
+            SessionListItem(
+                session_id=r["id"],
+                created_at=r["created_at"],
+                resume_claim=r["resume_claim"],
+                is_complete=bool(r["is_complete"]),
+                turn_count=r["turn_count"],
+            )
+            for r in rows
+        ]
+
+
+async def get_session_row(session_id: UUID) -> aiosqlite.Row | None:
+    async with connect() as db:
+        db.row_factory = aiosqlite.Row
+        cursor = await db.execute("SELECT * FROM sessions WHERE id = ?", (str(session_id),))
+        return await cursor.fetchone()
+
+
+async def get_transcript(session_id: UUID) -> list[TranscriptSegment]:
+    async with connect() as db:
+        db.row_factory = aiosqlite.Row
+        cursor = await db.execute(
+            "SELECT turn_number, speaker, text FROM transcript_segments "
+            "WHERE session_id = ? ORDER BY id ASC",
+            (str(session_id),),
+        )
+        rows = await cursor.fetchall()
+        return [TranscriptSegment(turn_number=r["turn_number"], speaker=r["speaker"], text=r["text"]) for r in rows]
+
+
+async def get_evidence(session_id: UUID) -> list[EvidenceRecord]:
+    async with connect() as db:
+        db.row_factory = aiosqlite.Row
+        cursor = await db.execute(
+            "SELECT competency, node, status, note, turn_number FROM evidence "
+            "WHERE session_id = ? ORDER BY id ASC",
+            (str(session_id),),
+        )
+        rows = await cursor.fetchall()
+        return [
+            EvidenceRecord(
+                competency=r["competency"],
+                node=r["node"],
+                status=r["status"],
+                note=r["note"],
+                turn_number=r["turn_number"],
+            )
+            for r in rows
+        ]

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/evidence.py

@@ -0,0 +1,48 @@
+"""The competency/evidence graph (spec §21) for this slice's one demo
+competency: RAG Architecture. A production system would load this from a
+job-description-driven competency library; here it's a fixed definition
+matching the §36 demo scenario exactly.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+
+from .models import EvidenceStatus
+
+COMPETENCY = "rag_architecture"
+
+NODES: dict[str, str] = {
+    "retrieval_design": "How retrieval is designed — what's fetched vs. handled directly by the model",
+    "relevance_handling": "What happens when retrieval returns irrelevant or low-quality documents",
+    "production_scaling": "How the system behaves and is investigated under a large traffic/latency increase",
+}
+
+
+@dataclass
+class NodeState:
+    status: EvidenceStatus = "not_addressed"
+    note: str = ""
+
+
+@dataclass
+class EvidenceGraphState:
+    nodes: dict[str, NodeState] = field(default_factory=lambda: {n: NodeState() for n in NODES})
+
+    def apply(self, node: str, status: EvidenceStatus, note: str) -> None:
+        if node not in self.nodes:
+            # The LLM invented a node id we don't track — ignore rather than
+            # crash the turn; the raw note is still visible in the transcript.
+            return
+        self.nodes[node] = NodeState(status=status, note=note)
+
+    def sufficiently_covered(self) -> bool:
+        addressed = [n for n in self.nodes.values() if n.status != "not_addressed"]
+        return len(addressed) == len(self.nodes)
+
+    def summary_for_prompt(self) -> str:
+        lines = []
+        for node_id, description in NODES.items():
+            state = self.nodes[node_id]
+            lines.append(f"- {node_id} ({description}): {state.status}" + (f" — {state.note}" if state.note else ""))
+        return "\n".join(lines)

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/llm.py

@@ -0,0 +1,32 @@
+"""LLM provider (spec §31: interchangeable behind a narrow interface).
+Only implementation right now is local Ollama; ASR/TTS have their own
+similarly narrow wrappers in asr.py/tts.py. Nothing outside this module
+imports `ollama` directly.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+
+import ollama
+
+from .models import AgentTurnDecision
+
+MODEL = os.environ.get("OLLAMA_MODEL", "qwen2.5:7b-instruct")
+
+_client = ollama.AsyncClient()
+
+
+async def decide_turn(system_prompt: str, user_prompt: str) -> AgentTurnDecision:
+    response = await _client.chat(
+        model=MODEL,
+        messages=[
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": user_prompt},
+        ],
+        format=AgentTurnDecision.model_json_schema(),
+        options={"temperature": 0.4},
+    )
+    content = response.message.content
+    return AgentTurnDecision.model_validate(json.loads(content))

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/main.py

@@ -0,0 +1,95 @@
+from __future__ import annotations
+
+from contextlib import asynccontextmanager
+from uuid import UUID
+
+from fastapi import FastAPI, HTTPException, Request
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import StreamingResponse
+
+from . import agent, asr, db, tts
+from .models import (
+    SessionListItem,
+    SessionSummary,
+    StartSessionRequest,
+    StartSessionResponse,
+    SynthesizeRequest,
+    TranscribeResponse,
+    TurnRequest,
+    TurnResponse,
+)
+
+
+@asynccontextmanager
+async def lifespan(_app: FastAPI):
+    await db.init_db()
+    yield
+
+
+app = FastAPI(title="PyInterviewBot ai-service", lifespan=lifespan)
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+
+@app.get("/health")
+async def health() -> dict[str, str]:
+    return {"status": "ok"}
+
+
+@app.post("/sessions", response_model=StartSessionResponse)
+async def create_session(_req: StartSessionRequest) -> StartSessionResponse:
+    session_id, question_text = await agent.start_session()
+    return StartSessionResponse(session_id=session_id, question_text=question_text)
+
+
+@app.post("/sessions/{session_id}/turn", response_model=TurnResponse)
+async def turn(session_id: UUID, req: TurnRequest) -> TurnResponse:
+    try:
+        return await agent.take_turn(session_id, req.transcript)
+    except ValueError as exc:
+        raise HTTPException(status_code=404, detail=str(exc)) from exc
+
+
+@app.get("/sessions", response_model=list[SessionListItem])
+async def list_sessions() -> list[SessionListItem]:
+    return await db.list_sessions()
+
+
+@app.get("/sessions/{session_id}/summary", response_model=SessionSummary)
+async def summary(session_id: UUID) -> SessionSummary:
+    row = await db.get_session_row(session_id)
+    if row is None:
+        raise HTTPException(status_code=404, detail=f"unknown session {session_id}")
+    return SessionSummary(
+        session_id=session_id,
+        resume_claim=row["resume_claim"],
+        is_complete=bool(row["is_complete"]),
+        transcript=await db.get_transcript(session_id),
+        evidence=await db.get_evidence(session_id),
+    )
+
+
+@app.post("/asr/transcribe", response_model=TranscribeResponse)
+async def transcribe(request: Request) -> TranscribeResponse:
+    pcm16 = await request.body()
+    text = await asr.transcribe(pcm16)
+    return TranscribeResponse(transcript=text)
+
+
+@app.post("/tts/synthesize")
+async def synthesize(req: SynthesizeRequest) -> StreamingResponse:
+    return StreamingResponse(
+        tts.synthesize_stream(req.text),
+        media_type=f"audio/L16;rate={tts.SAMPLE_RATE};channels=1",
+    )
+
+
+def main() -> None:
+    import uvicorn
+
+    uvicorn.run("pyinterviewbot_ai.main:app", host="127.0.0.1", port=8000, reload=False)

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/models.py

@@ -0,0 +1,92 @@
+"""Pydantic schemas: the HTTP contract with the Rust gateway, plus the
+structured-output schema the LLM is constrained to when deciding a turn.
+"""
+
+from __future__ import annotations
+
+from typing import Literal
+from uuid import UUID
+
+from pydantic import BaseModel, Field
+
+EvidenceStatus = Literal["confirmed", "weak", "contradicted", "not_addressed"]
+NextAction = Literal["probe", "clarify", "challenge", "move_on", "simulate", "conclude"]
+
+
+class StartSessionRequest(BaseModel):
+    resume_seed: bool = True
+
+
+class StartSessionResponse(BaseModel):
+    session_id: UUID
+    question_text: str
+
+
+class TurnRequest(BaseModel):
+    transcript: str
+
+
+class TurnResponse(BaseModel):
+    ai_text: str
+    is_complete: bool
+
+
+class TranscribeResponse(BaseModel):
+    transcript: str
+
+
+class SynthesizeRequest(BaseModel):
+    text: str
+
+
+class EvidenceUpdate(BaseModel):
+    """One evidence-graph node update the LLM produces after hearing an answer."""
+
+    node: str = Field(description="Evidence node id this update applies to")
+    status: EvidenceStatus
+    note: str = Field(description="One sentence: what the candidate said that supports this status")
+
+
+class AgentTurnDecision(BaseModel):
+    """The full structured output the interview agent LLM call must produce
+    each turn. The gateway never sees this — only `question_text` (as
+    `ai_text`) and `is_complete` cross the HTTP boundary back to it.
+    """
+
+    evidence_updates: list[EvidenceUpdate]
+    next_action: NextAction
+    question_text: str
+    is_complete: bool
+
+
+class TranscriptSegment(BaseModel):
+    turn_number: int
+    speaker: Literal["ai", "candidate"]
+    text: str
+
+
+class EvidenceRecord(BaseModel):
+    competency: str
+    node: str
+    status: EvidenceStatus
+    note: str
+    turn_number: int
+
+
+class SessionListItem(BaseModel):
+    session_id: UUID
+    created_at: str
+    resume_claim: str
+    is_complete: bool
+    turn_count: int
+
+
+class SessionSummary(BaseModel):
+    """Recruiter-facing view — the one place all this detail is allowed to
+    surface (spec §24)."""
+
+    session_id: UUID
+    resume_claim: str
+    is_complete: bool
+    transcript: list[TranscriptSegment]
+    evidence: list[EvidenceRecord]

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/resume.py

@@ -0,0 +1,26 @@
+"""Resume claim seeding.
+
+This MVP slice does not implement resume upload or a real resume parser
+(spec §15 describes claim extraction from an uploaded document generally —
+that's out of scope here). Instead it seeds the one claim used by the §36
+demo scenario. A future resume parser would produce `ResumeClaim` objects
+the same shape as this one; nothing downstream needs to change.
+"""
+
+from dataclasses import dataclass
+
+
+@dataclass(frozen=True)
+class ResumeClaim:
+    competency: str
+    text: str
+
+
+SEED_CLAIM = ResumeClaim(
+    competency="rag_architecture",
+    text="Designed an enterprise RAG platform using Azure OpenAI, Azure AI Search and AKS.",
+)
+
+
+def seed_claim() -> ResumeClaim:
+    return SEED_CLAIM

pyinterviewbot_ai-0.1.0/src/pyinterviewbot_ai/tts.py

@@ -0,0 +1,65 @@
+"""TTS provider (spec §31): local Piper. Synthesizes and streams raw
+PCM16LE mono audio chunk-by-chunk as they're produced, rather than
+buffering the whole utterance — this is what lets the gateway start
+forwarding audio to the browser before the sentence has finished
+synthesizing (spec §27).
+"""
+
+from __future__ import annotations
+
+import asyncio
+import os
+import queue
+import threading
+from collections.abc import AsyncIterator
+
+from piper.voice import PiperVoice
+
+MODEL_PATH = os.environ.get("PIPER_MODEL_PATH", "models/en_US-lessac-medium.onnx")
+
+_voice: PiperVoice | None = None
+_lock = threading.Lock()
+
+SAMPLE_RATE = 22050  # must match the loaded voice's config; asserted at load time
+
+
+def _get_voice() -> PiperVoice:
+    global _voice
+    with _lock:
+        if _voice is None:
+            voice = PiperVoice.load(MODEL_PATH)
+            assert voice.config.sample_rate == SAMPLE_RATE, (
+                f"voice sample rate {voice.config.sample_rate} != expected {SAMPLE_RATE}; "
+                "update tts.SAMPLE_RATE and the candidate client's playback rate together"
+            )
+            _voice = voice
+    return _voice
+
+
+_SENTINEL = object()
+
+
+def _synthesize_worker(text: str, out_queue: "queue.Queue[object]") -> None:
+    try:
+        voice = _get_voice()
+        for chunk in voice.synthesize(text):
+            out_queue.put(chunk.audio_int16_bytes)
+    except Exception as exc:  # noqa: BLE001 — surfaced to the caller via the queue
+        out_queue.put(exc)
+    finally:
+        out_queue.put(_SENTINEL)
+
+
+async def synthesize_stream(text: str) -> AsyncIterator[bytes]:
+    out_queue: "queue.Queue[object]" = queue.Queue()
+    thread = threading.Thread(target=_synthesize_worker, args=(text, out_queue), daemon=True)
+    thread.start()
+
+    loop = asyncio.get_event_loop()
+    while True:
+        item = await loop.run_in_executor(None, out_queue.get)
+        if item is _SENTINEL:
+            break
+        if isinstance(item, Exception):
+            raise item
+        yield item  # type: ignore[misc]